Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                
0% found this document useful (0 votes)
130 views

2019 Issue-2

Uploaded by

Tao Hou
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
130 views

2019 Issue-2

Uploaded by

Tao Hou
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 84

Custom Actuator Solutions

After nearly four decades of global


success, Hitec continues to deliver
unrivaled reliability, power, speed, and
durability as a world leader in actuator
design. Hitec customers rely on our
engineering and manufacturing teams
to deliver affordable, precision solutions
for a wide range of applications.
Our off-the-shelf and purpose-built
custom products regularly incorporate
innovative features such as magnetic
encoders, brushless motors and
titanium gears, delivering optimal
performance for all your industrial
and commercial applications.

Hitec Commercial Solutions is


proud to offer extensive actuator
lines, custom services, and a
• Sizes Range from 10 - 33mm variety of unmanned platforms, all
• Contactless Position Sensors engineered to effortlessly conquer
• Wide Voltage Operation your most demanding challenges.
• High-Resolution Programmability
• CAN, uCAN & RS-485 Interfaces Available
• Customized Options Available

www.hitecnology.com

Actuator Engineering Reimagined


Hitec Commercial Solutions, LLC Booth No. 4005

9320 Hazard Way, Suite D, San Diego, CA 92123 | (858) 737-9220


Best service. Best Warranty. Best vaLue. no one Beats LDG! contact your Favorite DeaLer toDay!

Where to use LDG


BaLuns & ununs
Not sure which balun or unun is right for your antenna? See our handy chart
below to help you determine which is the best fit for your set-up. All LDG
$
30 ea.
baluns and ununs handle up to 200 Watts PEP and cover frequencies from 200 Watts PEP
1.8 to 30MHz. Visit us at www.ldgelectronics.com or see your favorite dealer
1.8-30MHz
today to learn more and to see our full line of products.

dipolE laddEr linE/TWin lEad


length = 468/freq

1:1 4:1
Balun Balun

Coax Coax

18' Thru 43' Ground End FEd WirE


VErTiCal planE 30' - 135'

9:1
unun
4:1
unun
Coax

Coax >30' Coax Cable

1:1 choke
Coax unun Coax

410-586-2177 support@ldgelectronics.com www.ldgelectronics.com


03/04.2019 VOLUME 17 • NO. 02

For more details on subscribing,


see our ad on Page 82.

Subscription Information
SERVO Magazine — PO Box 15277 Departments
North Hollywood, CA 91615-9218
Call 877-525-2539 or go to www.servomagazine.com 06 Mind/Iron
Subscribe • Gift • Renewal • Change of Info A Renaissance of the Kit
Building Era?
07
22 Bots in Brief 20
Bio-Feedback
New Products
69 SERVO Webstore
77 The GearBox
78 In the News
81 Advertiser’s Index

Combat Zone

• Guardian of the (Sea) • Make Some Noise


Galaxy • Excercise Coach,
• Bend It Like Nadia Robot Style 34 Vortex — The Shell Spinner
Adventure­­
• Get a (Soft) Grip • Locomotion Notion
• Independence Day • Kroger Really Delivers
• Poetry in Motion
• We Do Windows
• A No-Brainer
• Digits Deliver 38 The Bot of Theseus
• Fly, MetaFly, Fly
• Internet from the Sky
• To the (Jupiter’s)
Moon
42 I Build BattleBots
• Get Racey • Can You Feel It?
SERVO Magazine (ISSN 1546-0592/CDN Pub Agree#40702530) is published
• Help Me! I’m Melting! • Learning to Adapt bi-monthly for $26.95 per year by T & L Publications, Inc., 2279 Eagle Glen
• Man’s Best Friend Parkway, #112-481, Corona, CA 92883. Periodicals postage PAID at Corona, CA
and at additional entry mailing offices. POST­MASTER: Send address changes
to SERVO Magazine, P.O. Box 15277, North Hollywood, CA 91615 or Station A,
P.O. Box 54, Windsor ON N9A 6J5; cpcreturns@servomagazine.com.

4 SERVO 03/04.2019
Page 56

Page 14

Page 46

08 Developing Robotic Behaviors 56 Balancing a Robot on a Ball


by John Blankenship by John Blankenship
Developing a robotic behavior often involves activities such Assume you have a mobile robot sitting on a large ball.
as brainstorming, testing, evaluation, innovation, etc. One of Imagine programming the robot so that if the ball is
the biggest problems associated with such endeavors can be moved, the robot will climb to the top of the ball and level
determining why algorithms fail. This article explores these itself. If building such a robot excites you, this article will
activities while developing and perfecting a simple — yet demonstrate some key points to help get you started.
interesting — behavior using a simulator.
60 Giving Double Jeopardy an
14 Human Media Lab Takes LEGO to Unconstitutional Amount of Shots
the Next Level Twin Tweaks: Twin brothers hack whatever’s put
by Ryan Fleming in front of them, then tell you about it.
For more than half a century, LEGO building blocks have by Bryce and Evan Woolley
sparked the imaginations of both children and adults across Even though we consider our 2018 season on BattleBots a
the globe, helping to transform imaginative concepts into smashing success, we knew there were plenty of upgrades
concrete (or, more accurately, plastic) reality. Now, thanks that would help Double Jeopardy stand a better chance
to a collaboration between Queen’s University in Ontario, against the mechanized rampage of the opposing bots in
Canada and LEGO Creative Play Lab, LEGO’s bricks are the 2019 season. The most obvious area of improvement
finding their way into a new dimension. was the cannon.

46 Make a Spring-Loaded Pen 70 Service Robots are Knocking


Holder for Your CNC Router for $5! on Consumer’s Doors
by Roger Secura Then & Now: Advances in robotics from the past
A good CNC router pen holder can cost up to $75 or more. up through today.
If you’ve been thinking of buying one, then maybe you by Tom Carroll
should consider making one instead. In this article, I’ll show Look at some of the failures of once promising home service
you how to make a spring-loaded pen holder for five bucks robot businesses, and the different aspects and obstacles
— and thanks to some common plumbing and hardware of what it will take to produce a truly successful “home”
parts from Lowe’s and Home Depot, you can make one in automaton.
less than an hour.
Page 70
50 REVIEW: Vorpal Hexapod
by Carol Lynn Hazlett
What drew me to look at this robot was how cute it is. What
made me buy it is how versatile it is. Plus, it has a low price!
To let you know how much I like this robot, I bought five
of them! The Vorpal is a very versatile hexapod for STEM,
gaming, and lots of fun.
SERVO 03/04.2019 5
Mind / Iron
Published Bi-Monthly By
T & L Publications, Inc.
2279 Eagle Glen Pkwy #112-481
Corona, CA 92883
by Bryan Bergeron, Editor OFFICE: (951) 371-8497
FAX: (951) 371-3052

A Renaissance of the Kit Building Era?


WEBSTORE: (800) 783-4624
www.servomagazine.com

A
cornerstone of my SUBSCRIPTIONS
Toll Free: (877) 525-2539
early education in
Outside US: (818) 487-4545
electronics was a P.O. Box 15277
series of electronic kits from North Hollywood, CA 91615
the HeathKit company. I’m subscribe@servomagazine.com
not talking about a plastic
breadboard of flashing PUBLISHER
LEDs like the kits once sold Larry Lemieux
by RadioShack, but real publisher@servomagazine.com
electronics: Test equipment,
including oscilloscopes ASSOCIATE PUBLISHER/
and digital multimeters; ADVERTISING SALES
amateur radio transmitters, Robin Lemieux
Elecraft 500W linear amplifier under
transceivers, and amplifiers; robin@servomagazine.com
construction on my desk.
and, of course, the line
of HERO robots. I was smitten by the circuit to find the best impedance EXECUTIVE DIRECTOR/
HERO 1, and hooked on robotics for life. match between the transmitter stage PRODUCTION
Sean Lemieux
Although the HeathKit brand and the antenna system.
sean@servomagazine.com
has been resurrected, its offerings The Elecraft kits are top notch in
are limited to a handful of consumer- several aspects. One is price. These
EDITOR
oriented kits (www.heathkit.com). kits are at least as expensive as Bryan Bergeron
In the dwindling electronic kits market finished products from the big-name techedit-servo@yahoo.com
(which once included companies such manufacturers. For example, the
as Ramsey Electronics), at least one Elecraft Panadapter alone costs as CONTRIBUTING EDITORS
company has retained and perhaps much as one of my Icom IC-7300s — a Tom Carroll Kevin Berry
evolved the DNA of the original complete transceiver that contains an Bryce Woolley Evan Woolley
HeathKit. integrated panadapter. John Blankenship Carol Hazlett
I’m talking about Elecraft (www. The other consideration is quality Ryan Fleming AJ McNamara
elecraft.com), founded in 1998 of build. There’s no skimping with Stephen Pasek Andrea Gellatly
Roger Secura
by two amateur radio enthusiasts. these kits. They may cost as much
The company offers a variety of or more as similar equipment from
WEBSTORE MARKETING/
sophisticated high-end kits for test the big-name companies, and there’s
equipment and amateur radio. There good reason. The components are top COVER GRAPHICS
Brian Kirkpatrick
are no LED clocks sold here. notch, and each board is fully tested brian@servomagazine.com
I’ve purchased and built several and burned in before shipping out to
of the Elecraft kits to-date: a 100 watt customers. Copyright © 2019 by T & L Publications, Inc.
transceiver/general coverage receiver Finally, there’s the nature of the All Rights Reserved
(K3S) that covers two meters as well kits. With a few exceptions, the kits All advertising is subject to publisher’s approval.
as the high frequency (HF) amateur are modular for both expansion and We are not responsible for mistakes, misprints,
or typographical errors. SERVO Magazine
radio bands; a spectrum analyzer upgrade. Take the K3S transceiver. The
assumes no responsibility for the availability
(Panadapter) that provides a real time amplifier, power supply, and automatic or condition of advertised items or for the
view of activity on the radio spectrum antenna tuner boards are separate and honesty of the advertiser. The publisher makes
and view of the transmitted waveform; independently upgradeable. no claims for the legality of any item advertised
a 500W linear amplifier (see photo); I could have purchased the in SERVO. This is the sole responsibility of the
and a 500W automatic antenna tuner. transceiver with, for example, a 10 advertiser. Advertisers and their agencies agree
to indemnify and protect the publisher from
There are no robots offered — watt power amplifier board, which I
any and all claims, action, or expense arising
although you could argue that the then could have upgraded to 100W. from advertising placed in SERVO. Please send all
automatic antenna tuner performs Instead, I purchased the kit with the editorial correspondence, UPS, overnight mail,
a dangerous, dirty, and dull job of 100W amplifier option. Similarly, and artwork to: 2279 Eagle Glen Pkwy #112-
switching inductors in and out of if I had purchased an earlier K3S 481, Corona, CA 92883.

6 SERVO 03/04.2019
transceiver, I could have made the would simply swap out a board and screwdrivers are all you need to
board swaps to bring the system on download new firmware. Quality of completely tear down a device.
par with the current K3S model. It’s build and maintainability aside, I’d still I was never into the “black box”
this modular approach to construction opt for a kit over a ready-made device. approach to electronics. I have to
and maintenance that I’d like to relate It isn’t about cost-savings. Elecraft know what’s inside. I believe there
to the SERVO community. It would offers prebuilt and fully tested units for are enough robotics and electronics
be an excellent approach to robot $100-$200 more than their kits, but enthusiasts that share the same need
building to ensure longevity of the I’ll have none of it. It’s about spending to know to keep the kit building
product and to ease the burden of tens of hours getting to know the market alive.
redesign. insides of a transceiver or amplifier. For that reason, I’ll do what I
When it’s time to upgrade a Of knowing how to get to the internal can to support the cause. I’m already
classroom robot to address the needs fuses and the power transistors, thinking about my next kit. How about
of a more advanced audience, one and that two sizes of Phillips head you? SV

BioFeedback
BASIC-ly Impressed
I’ve been very impressed with
the latest set of articles by John
Blankenship. I followed his series on
RobotBASIC and the S3 Scribbler
robot, and learned quite a bit about
the capabilities of the hardware and
software.
Now, his latest series demonstrates
a capability of RobotBASIC that
communicates with and controls many
different hardware platforms. He does
so by deftly introducing the concept
of TCP/IP (and UDP) along with ports.
The RobotBASIC programs provided
are clear and create a foundation for
further experimentation.
The introduction of the ECHO
(Alexa) device was a welcome surprise.
I didn’t know that Alexa had the
RobotBASIC “Skill.” So, I would hope
to see many new articles on combining
the network communications features
of RB’s TCP and the voice recognition
of Alexa. These form a truly “hardware
agnostic” control platform that will be
of interest to many SERVO subscribers.
Rick Wagoner

Thanks for the feedback, Rick.


Glad you’re enjoying my articles!
The TCP and UDP communication
(without additional hardware) is a
great feature of RobotBASIC. There
are several free downloads on the
FAQ tab at RobotBASIC.org that
provide extensive information on
Internet communication. As always,
let me know if you have questions or
problems.
John Blankenship
SERVO 03/04.2019 7
Developing
Robotic
Behaviors
By John Blankenship

Developing a robotic behavior often involves


activities such as brainstorming, testing,
evaluation, innovation, etc. One of the biggest
problems associated with such endeavors can
be determining why algorithms fail. This article
explores these activities while developing
and perfecting a simple — yet interesting —
behavior using a simulator.
8 SERVO 03/04.2019
O
ne of the joys of hobby robotics is finding ways
to make a robot solve problems and accomplish main:
goals in — at least what appears to be — an gosub Obstacles FIGURE 1.
gosub InitRobot
intelligent manner. In order to address the gosub Explore
development process for a robotic behavior in an end
article format, we need a problem that is simple
enough to be explored in a few pages, yet complex enough Explore:
to need a trial-and-error approach. After some thought, an while true
appropriate project came to mind. while not rFeel()
Assume our robot’s environment is a cluttered room rForward 1
wend
and our goal is to have the robot explore that room in a amount = -45+random(90)
manner that makes it venture into all areas of the room on a if random(100)<50
fairly equal basis. dir = -1
This is a much different behavior than a robot just else
wondering randomly while avoiding objects. dir = 1
One easy way of viewing this difference is to utilize endif
a simulation. RobotBASIC will be used since it’s a free rTurn dir*amount
wend
language that includes an integrated robot simulator return
(download from www.RobotBASIC.org).
Figure 1 shows a partial program that moves the InitRobot:
simulated robot randomly through a cluttered room. Figure rLocate 200,400,180
2 shows the environment and the path taken by the robot rInvisible RED
over a short period of exploration. Of course, since the rPen Down
robot’s behavior is randomized, the path taken will be return
different each time the program is run. Obstacles:
This example is typical though, in that the robot LineWidth 5
can easily get stuck in a partially enclosed region of the Rectangle 0,0,799,599
environment for an extended period of time. Even if the LineWidth 1
robot escapes (as it finally did in Figure 2), it’s unlikely Circle 250,250,375,500,black,black
that the robot will explore all of its environment in any circle 480,100,700,400,black,black
reasonable time. rectangle 700,350,800,400,black,black
rectangle 20,250,150,400,black,black
Simulations are very helpful for developing algorithms. return
If a real robot was used to perform this procedure, not
only would you have to create a suitable
test environment, but the time taken for a
physical robot to move around can be far
longer than a simulated one, which can
greatly extend the development process. FIGURE 2.
The important point though, is that we
need a much better algorithm for controlling
the robot’s behavior.
Before improving our code, let’s examine
the algorithm in the Explore module of
Figure 1.
The Explore module consists of an
endless loop that moves the robot forward
as long as none of the robot’s perimeter
proximity sensors (accessed by the simulator’s
rFeel function) detect a nearby object.
If something is detected, the robot turns
to a random heading between -45° and +45°.
While this does move the robot around its
environment, the quality of its movements
(as we have seen) is not ideal.
SERVO 03/04.2019 9
Obviously, a totally random movement is
not a good choice.

Improving the Original


Algorithm
One simple way of improving the
algorithm would be to change the size of the
random turns the robot makes. This could
easily be tested with the simulation to try to
find acceptable parameters before testing the
program with a real robot.
Unfortunately, even if you find
parameters that work reasonably well in
one environment, they will probably fail in a
different one.
That’s the nice thing about simulation.
You can easily test a variety of environments
with different parameters to see if you can
find something that will always work.
Let’s try to improve on this program
FIGURE 3. using a different approach. Before reading
further, consider this problem and think
about how you might change the robot’s
Explore: behavior to get better results.
while true
// use only three rFeel Sensors
while not (rFeel()&14) A Better Algorithm
rForward 1 One option is to create an entirely new
wend algorithm. Instead of just making the robot make
gosub TurnToOpenArea completely random turns, we need a way for it to
wend turn to a more reasonable heading. Such a decision
return
FIGURE 4. could be made in many ways.
One simple idea that might work would be to
TurnToOpenArea:
have the robot look for the largest open area near its
gosub FindLeftRightDist
if RtDist>LeftDist current heading.
rTurn 50 // right is more open Before we start developing this algorithm on a
else simulator, we should consider the real robot we will
rTurn -50 // left is more open eventually want to use.
endif Since we’re using RobotBASIC to develop the
return algorithm, it makes sense that we use a robot
capable of executing the same programs used for
FindLeftRightDist: the simulator. The RROS Tab on the RobotBASIC
LeftDist = 1000 webpage offers a free 200 page download to help
for a=-35 to -65 step 5 you build RobotBASIC compatible robots. This topic
if rRange(a)<LeftDist then was also addressed in the January 2018 issue of
LeftDist=rRange(a) SERVO in an article titled “RobotBASIC Robots for
next Beginners.”
RtDist = 1000 Additionally, the March 2018 issue discussed
for a=35 to 65 step 5 how the Parallax Scribbler S3 could be expanded and
if rRange(a)<RtDist then RtDist=rRange(a) controlled with RobotBASIC. For purposes of this
next
article, we will assume the algorithms will eventually
return
be utilized on a Scribbler S3 as shown in Figure 3.
10 SERVO 03/04.2019
TurnToOpenArea:
repeat
gosub FindLeftRightDist FIGURE 5.
// turn away from a
// stuck situation
if (LeftDist<50
and RtDist<50) or
(random(100)<5)
rTurn -90+random(180)
endif
until LeftDist>50 or
RtDist>50
if RtDist>LeftDist
rTurn 50
// right is more open
else
rTurn -50
// left is more open
endif
return FIGURE 6.
FIGURE 7.
The modified S3 robot utilizes PING)))
ranging sensors which have a much wider
beam than the simulated IR rangers on the
RobotBASIC simulation. In addition, the S3
only has three rFeel sensors compared to five
on the simulator.
For these reasons, additional code will
be used to make the simulation act more like
the S3 to help ensure that the final algorithm
developed on the simulation will control the
real robot in a similar manner.

Implementing the New


Algorithm
Figure 4 shows a new version of
the Explore module and the subroutines
TurnToOpenArea and FindLeftRightDist.
Together, these modules use the ranging sensors to turn the Doing so reveals that the robot often sticks in a corner
robot either left or right depending on which area appears when the robot wants to turn towards an open area, when
to be the most open (longest diagonal range reading). both the left and right ranging distances are small.
We can fix this problem by having the robot turn to
Analyzing the Problem a random heading between -90° and +90° whenever the
distances detected by both rangers is small (less than 50 in
Unfortunately, this algorithm only works for a short this example). The program should also initiate this random
period of time before the robot gets stuck as shown in turn about 5% of the time regardless of the sensor readings
Figure 5. The reason for this problem is easy to find on to further ensure that the robot does not get stuck in an
the simulation by stepping through the code or by having unusual situation.
the program display the values of the left and right rangers The code to implement this is shown in the new version
when the problem occurs. of TurnToOpenArea in Figure 6.
SERVO 03/04.2019 11
Explore:
while true
// use only three rFeel
// Sensors
while not (rFeel()&14)
gosub FindLeftRightDist
if LeftDist<RtDist
rTurn 1
else
rTurn -1
endif
rForward 1
wend
gosub TurnToOpenArea
wend
return FIGURE 8. FIGURE 9.

Unfortunately, even though the new code of Figure


6 prevents the robot from getting stuck, the robot’s
Creating More Natural
movements (see Figure 7) don’t really seem very natural Movements
(when comparing them to how a human might move). As
you can see from the figure, the robot always moves in a One idea for creating a more natural movement would
straight line and doesn’t seem to be moving with any real be to make the robot try to remain in open areas as it moves
intent. (not just when it collides with something as in the previous
program).
Explore: This can be done by having the robot (while moving)
while true constantly check to see if the area is more open to the left
// use only three rFeel Sensors or right. This can be done as shown in Figure 8.
while not (rFeel()&14) The new code in Figure 8 simply checks to see if the
gosub FindLeftRightDist range distance to the left or right is larger, and then turns
// turn away half the time slowly in that direction while still moving forward.
This simple change causes the robot to always turn
if random(100)<50
slightly toward open spaces as shown in Figure 9. This new
d = 1
behavior certainly makes the robot seem to move more
else
intelligently.
d=-1
endif
if random(100)<50 Problems Still Exist
if LeftDist<RtDist Unfortunately, even though the robot of Figure 9
rTurn d*(1+random(3)) seems to have found its way to all areas of the room (in
else this running of the program) with more natural-looking
rTurn -d*(1+random(3)) movements, the robot gets stuck in a pattern moving
endif between the upper left and lower right corners.
endif This can be corrected by giving the robot a little curiosity
rForward 1 by adding a little randomness to its exploratory movements.
FIGURE 10. Figure 10 shows a new version of the Explore
wend
gosub TurnToOpenArea subroutine that adds such randomness. During normal
wend movement, it causes the robot to turn away from (rather
than toward) open spaces 50% of the time.
return
It also turns a random amount as it moves in the
12 SERVO 03/04.2019
exploratory mode. This produces the motion FIGURE 11.
shown in Figure 11.
Overall, the robot now explores all areas
of its environment on a fairly equal basis. It
does seem to move toward open spaces, but
not always. There is just enough randomness
to ensure that all areas of the environment
are visited.
If a robot used this algorithm to search
its environment for a dime on the floor,
it could certainly find it (as long as it had
sensors that could detect the dime).

Conclusion
A variety of programming changes
have shown how a robotic problem can be
analyzed and developed using a simulation.
The final program solves the stated
problem and can be used to control the
Scribbler S3. For information on using the
program with the S3, refer to the March 2018 SERVO article upon these algorithms and/or to create and develop robotic
mentioned earlier. Readers are encouraged to download behaviors of their own.
RobotBASIC
and implement Beware, though, a simulator can make testing and
their own ideas developing algorithms for robotic behaviors so fast and easy
to improve that you might find it addictive. SV

To post comments on this article and find any associated files and/or
downloads, go to www.servomagazine.com/magazine/issue/2019/02.

Make Precision
Metal Projects
User-friendly, budget-friendly,
making-things-happen-friendly.

The compact Tormach PCNC 440


runs on single-phase power and
fits just about anywhere.

Delivered* and ready to rumble


for under $8700.

Get started at tormach.com/440.

*Standard shipping included in the continental USA.

TORMACH.COM/440

SERVO 03/04.2019 13
Human Media
Lab Takes LEGO
to the Next Level
Using Motion Capture Cameras, Queen’s
University’s Human Media Lab Helps Create
Flying LEGO Bricks that Could Mimic Real
World Movements By Ryan Fleming
14 SERVO 03/04.2019
User interacting with a drone swarm using the butterfly-shaped embodied controller.

For more than half a century, LEGO building blocks have sparked the
imaginations of both children and adults across the globe, helping to
transform imaginative concepts into concrete (or, more accurately,
plastic) reality. Now, thanks to a collaboration between Queen’s
University in Ontario, Canada and LEGO Creative Play Lab, LEGO’s
bricks are finding their way into a new dimension.

B
y applying passive markers to colored LEGO bricks, several small robots all at once, but it’s another to have
researchers at Queen’s Human Media Lab (HML) them move in unison — that requires precision tracking.
are able to create a flying robotic swarm capable of To record drones in flight, the team at HML currently
mimicking the movement and placement of specific uses Vicon T-series motion capture cameras capable of
LEGO bricks. To demonstrate this, HML created a tracking dozens of moving objects at once at 120 frames
flexible gyroscopic controller that looks like a butterfly. Users per second. These cameras can isolate and record the
could then place red, green, or blue bricks on the butterfly movement of each drone without issue, then send that data
controller, and small quadcopter drones with matching to Vicon’s Tracker 3 software where it can be controlled
colors flew into the air in a formation exactly matching the using a peripheral device — even a LEGO controller.
layout on the controller. Users could then turn, flip, or bend HML showed off its cool new LEGO-infused project last
the controller and the drones would mimic the movements year at the LEGO World Expo in Copenhagen, Denmark,
in real time. where attendees — young and old — took to the skies
Swarm robotics is still a relatively young field in the with their designs. While this might be one of the most
world of robotics, in part due to the technology needed sophisticated uses of LEGO bricks to-date, the brick-maker is
to achieve an accurate swarm. It’s one thing to command no stranger to robotics.
SERVO 03/04.2019 15
In 1998, LEGO introduced the
LEGO MINDSTORMS line and also
A BitDrone flying LEGO block.
“intelligent bricks” that are capable
of receiving basic programming that
allowed them to perform simple task-
oriented movements. The product line
included a brick computer that controls
the system. It worked with a set of
modular sensors and motors, along
with LEGO “Technic” bricks designed to
build mechanical systems.
LEGO’s robotics endeavor created a
relatively small but passionate fanbase,
leading to global competitions and
ushering in a new scientific-focused
side to the massive LEGO brand.
As youth robotics programs have
proliferated, competitions like FIRST
LEGO League — where young builders
create LEGO-based solutions to real
world problems — have kept pace.
Given LEGO’s history with robotic
systems, it made the partnership with
HML a natural fit.
Queen’s University’s HML is known
for its groundbreaking work developing
flexible displays, having created
interactive displays (read: touchscreens)
on non-flat surfaces like soda cans
and even bendable tablets. The

Butterfly-shaped embodied controller with LEGO


brick decorations.

16 SERVO 03/04.2019
User interacting with
BitDrones in a grid,
creating a non-uniform
rational basis spline
(NURBS) surface.

collaboration with LEGO is an extension of HML’s BitDrone works fine for demonstrations using LEGO bricks. The goal is
system, which initially debuted in 2015. to make them smaller, so that the swarm can act like a form
HML’s BitDrones project began as a 3D modeling tool. of interconnecting programmable matter.
Programmers used it to model several unique scenarios, In a similar vein, HML aims to make the system work
including one where you could navigate a file folder using with far bigger drone swarms. While users at the LEGO
touchpads on drones. Each drone contained a digital World Expo controlled no more than 10 or 15 drones at
readout to represent files or subfolders, and the Vicon once, the team at Queen’s is hoping to push that number
system recorded the movements. LEGO bricks weren’t into the hundreds, or even the thousands.
yet in the equation, but even then, HML Director Dr. Roel Although amazing crowds is a nice bonus, the
Vertegaal acknowledged the system’s resemblance to the program’s potential reaches in multiple directions, opening
building blocks. up new possibilities for students and professionals alike.
“They sort of resemble flying LEGO bricks, but the For instance, “DisplayDrones” — one of three BitDrones
application goes beyond simple toys,” said Vertegaal. revealed back at the 2015 ACM Symposium for User
Currently, HML’s BitDrones are relatively large, which Interface Software and Technology — are fitted with high-

SERVO 03/04.2019 17
resolution cameras of their own,
User interacting with three LEGO drones.
allowing users to control the drones
with simple head movements while
teleconferencing. With this tech,
prospective buyers could remotely
inspect a home, business partners
could take a self-guided tour of a model
property, and much more.
Vertegaal also believes the
technology could help revolutionize
the ways in which we teach physics to
children.
“We believe [the BitDrones project]
has the potential to take experiential
learning to an entirely new level,”
said Vertegaal. “With this technology,
we are able to simulate the physics
of the natural world like gravity,
planetary orbits, and more, giving
children a chance to see what they
have long learned from textbooks and
two-dimensional depictions, in a real
physical environment.” SV

To post comments on this article


and find any associated files
and/or downloads, go to www.
servomagazine.com/magazine/
issue/2019/02.

Flock of BitDrones.

18 SERVO 03/04.2019
NEW PRODUCTS
Modern Build System ranges from $1.19-$6.49.

go BILDA is a modern build system designed for


engineers, artists, students, and makers. It’s metric,
lightweight, and has an ever-growing library of parts.
Low-Side U-Channel
Low-side
U-channel
Rapidly prototype like a pro or enter a educational provides
competition with the ability to make quick in-the-field similar
modifications using only a hex key. structural
From small tabletop robots to projects you can ride, integrity as the
goBILDA’s interstitial pattern scales like a champ. You don’t 1120 series
need a machine shop to make high precision projects. U-channel
goBILDA is ball bearing based and is manufactured with while saving
industry leading tolerances. New products include the on real estate.
following. The reduced height of the sides allows more compact
building while still providing three sides to build off of. It’s
goTUBE an excellent choice when building a chassis with motors
goTUBE is designed with a 32 mm OD in order to run in running perpendicular to the direction of the channel.
bearings and pillow blocks. The 16 mm pattern is threaded Pricing ranges from $1.39-$35.69.
into the ends of the tube, so you can easily mount drive For further information, contact:
components
to the tube goBILDA
for a no-slip www.gobilda.com
transmission of
power. The bore
of the tube is
both 14 mm in
Three-Axis Compact Piezo Flexure
diameter and 12 Nano-Positioner
mm square. The
14 mm diameter
allows bearings
to slide in the
B ased on a parallel-kinematic design with only one
lightweight moving platform for all three axes, PI’s P-616
is available with high precision and dynamics in a compact
ends. The 12 mm square bore caters to creating linear drives package. Also known as the NanoCube®, it is a small and
with lead screws. Pricing ranges from $2.99-$13.99. light system with capacitive feedback delivering 100 µm
travel range in three degrees of freedom.
Square Beams The piezo scanner can be combined with compact
Square beams may be one of the most versatile motorized micro-positioning stages (such as PI’s L-505
structural elements on the 8 mm linear pattern. Thru-holes miniature stage) to achieve long travel ranges up to 1” in
from every side a very small package, with combinations for coarse/fine
allow screws to alignment.
pass through The P-616 is operated with PI’s E-727 digital servo piezo
the beams and controller which offers the advantage of higher linearity,
tie into other faster settling, and scanning speed, and easy access to all
components. motion parameters compared to conventional analog piezo
Each end is controllers. An all-inclusive software package includes drivers
threaded so that for LabVIEW, dynamic libraries for Windows and Linux,
connecting one and MATLAB. Interfaces consist of USB, SPI, RS-232, and
square beam analog. Supported functions include a wave generator, data
to another only recorder, auto zero, and trigger I/O.
requires a single The XYZ piezo stage is driven by ceramic-encapsulated
screw. Whether preloaded and flexure motion-amplified PICMA® piezo
they get stacked side by side, end to end, or fastened actuators that provide better performance and reliability
perpendicular to one another, they’ll remain on grid. Pricing than conventionally insulated piezo actuators. Actuators,
20 SERVO 03/04.2019
guidance, and sensors are maintenance-free, not
subject to wear, and offer extraordinary lifetime
and cost-effectiveness. In reliability tests carried out
by NASA/JPL for the Mars Mission, the actuators
survived 100 billion cycles without failures.
For further information, contact:

PI
www.pi-usa.us/en

Since this product is a carrier board or breakout


board for the STSPIN220 low-voltage stepper motor
driver from STMicro’s, Pololu recommends careful
reading of the STSPIN220 datasheet (1MB pdf)
before using this product. Some of the driver’s key
features include:

• Simple step and direction control interface.


• Nine different step resolutions down to 256
microsteps: full-step, half-step, 1/4-step, 1/8-step,
1/16-step, 1/32-step, 1/64-step, 1/128-step, and
1/256-step.
• Adjustable current control lets you set the
maximum current output, which lets you use
voltages above your stepper motor’s rated voltage
to achieve higher step rates.
• Motor supply voltage: 1.8V to 10V (for a higher
voltage alternative, consider the STSPIN820
carrier, which operates from 7V to 45V).
• Can deliver 1.1A per phase continuously without
additional cooling.
• Can interface directly with 3.3V and 5V systems.
• Over-temperature thermal shutdown, over-current
Low-Voltage Stepper Motor Driver shutdown, and short circuit protection.
Carrier • Four-layer 2 oz copper PCB for improved heat
dissipation.

T he latest version of Pololu’s STSPIN220 Low-Voltage


Stepper Motor Driver Carrier ships with male header
pins installed, so no soldering is required to use it with an
• Exposed solderable ground pad below the driver
IC on the bottom of the PCB.
• Module size, pinout, and interface match those
appropriate 16-pin socket or solderless breadboard. of Pololu’s A4988 stepper motor driver carriers in
This breakout board for STMicro’s STSPIN220 low- most respects.
voltage microstepping bipolar stepper motor driver offers
microstepping down to 1/256 of a step and operates from This product ships with all surface-mount components,
1.8V to 10V, allowing stepper motors to be powered with including the STSPIN220 driver IC. Price is $5.95, with
voltages that are too low for other drivers. It can deliver up quantity pricing available.
to approximately 1.1A per phase continuously without a For further information, contact:
heatsink or forced air flow (up to 1.3A peak). The module
has a pinout and interface that are very similar to that
of Pololu’s A4988 carriers, so it can be used as a drop-in
Pololu
replacement for those boards in many applications. www.pololu.com
SERVO 03/04.2019 21
bots
IN BRIEF
Guardian of the (Sea)
Galaxy
I t’s always cool to see lionfish
while snorkeling or scuba diving;
they’re truly spectacular looking.
Since they’re covered in flamboyant
spines, they feel safe enough to let
you get close. Unfortunately, lionfish
don’t make for very good oceanic
neighbors, and in places where
they’re an invasive species and have
few native predators (like most of
the Atlantic coast of the United
States), they basically eat anything
that moves while breeding almost
continuously.
The trouble with this is where
many lionfish hang out and breed, it’s too deep for most recreational divers. Enter Robots in Service of the
Environment (RSE).
Founded by Colin Angle of iRobot, RSE has been developing remote controlled underwater robots that can
efficiently hunt, stun, and capture lionfish at depths of 400 feet for up to 60 minutes at a time, and they’ve now
introduced their third-generation lionfish-hunting robot.
The RSE Guardian LF1 Mark 3 prototype is a modular unmanned underwater vehicle (UUV) designed
specifically for lionfish hunting. Two paddles at the front of the robot can zap any unsuspecting victim that swims
between them, stunning it instantly. The incapacitated fish is then sucked into the body of the robot. Up to 10 fish
can be captured before the robot needs to return to the surface.
The battery, sensing, propulsion, and nearly everything else is contained in modules bolted to the chassis
of the robot, making repairs and upgrades simple. Control is fully manual at the moment and dependent on
a tether to the
surface, although
RSE is exploring how
assistive autonomy can
make piloting easier.

The RSE Guardian LF1 Mark 3 UUV has a


modular design featuring: 1. Stunning panels;
2. Capture system; 3. Power enclosure; 4.
Main ROV enclosure; 5. Surface tether; 6.
6-DOF navigation system; 7. Stun enclosure;
and 8. Fish retention tines.

Images courtesy of RSE.

22 SERVO 03/04.2019
bots
IN BRIEF

Nadia 3D-printed
plastic mockup
in different poses
(clockwise from left):
foot up, stretching,
kneeling, curled up in
a ball, and sitting.
Images courtesy of
IHMC.

Bend It Like Nadia originally developed in collaboration with IIT for the HyQ

G etting robots to do the same kinds of things that humans


can is quite a challenge — especially since it requires
getting complex hardware and software to work together.
quadruped.
The ISAs are 3D printed out of titanium and completely
self-contained. They have a power-to-weight ratio substantially
Shown in the photo is a 3D-printed mockup of a new better than human muscles (if you don’t count the rest of the
humanoid from the robotics group at the Institute for Human hydraulic system), but they are still bulky enough that designing
& Machine Cognition (IHMC) in joints around them can be tricky.
Pensacola, FL. It’s named Nadia, after the It’s especially tricky since Nadia (a
famous gymnast Nadia Comâneci, and gymnast, remember) needs to have the
is being funded by the Office of Naval same range of motion that a human
Research (ONR). does, with the ability to do things
The mockup contains almost all like squat, crawl, curl up into a ball,
of the components that will be on the take a yoga class, or whatever else is
real robot (the motorcycle helmet necessary.
may or may not be the robot’s real IHMC has an enormous amount
head). Working with the design this of experience with walking robots.
way helps IHMC make sure everything They came in second at the DARPA
fits together properly and that the Robotics Challenge with their Running
actuators and joints provide the Man Atlas (one of just three teams
flexibility and range of motion they’re to score a perfect 8 out of 8) and
looking for. they’ve continued to advance bipedal
Nadia will be powered by locomotion using both Atlas and
hydraulics using Moog’s Integrated NASA’s Valkyrie.
Smart Actuators (ISAs). These were
SERVO 03/04.2019 23
Get a (Soft) Grip

T o give soft robots a bit of a


hand, researchers from MIT
and Harvard have developed a
new gripper that’s both supple
and strong: a cone-shaped origami
structure that collapses in on
objects much like a Venus flytrap,
to pick up items that are as much
as 100 times its weight. This
motion lets the gripper grasp
a much wider range of objects
such as soup cans, hammers, wine
glasses, drones, and even a single
broccoli floret.
The structure of this new
gripper takes an entirely different
form. Cone-shaped, hollow, and
vacuum-powered, the device was
inspired by the “origami magic
ball” and can envelope an entire
object and successfully pick it up.
The gripper has three parts:
the origami-based skeleton structure; the airtight skin to production lines in factories. Not surprisingly, the shape of the
encase the structure; and the connector. gripper makes it more difficult to grasp something flat, like a
The team created it using a mechanical rubber mold sandwich or book.
and a special heat-shrinking plastic that self-folds at high In the future, the team hopes to try to solve the problem
temperatures. of angle and orientation by adding computer vision that would
The robot currently works best with cylindrical objects let the gripper “see,” therefore, making it possible to grasp
like bottles or cans, which could someday make it an asset for specific parts of objects.

Independence Day
A ssistive technologies can benefit people with disabilities by giving them greater independence and control over
their lives, like the assistant robot developed by a team at Georgia Tech to help people who have severe motor
impairments.
“Our goal is to give people with limited use of their own bodies access to robotic bodies so they can interact with
the world in new ways,” Professor Charlie Kemp from
the Wallace H. Coulter Department of Biomedical
Engineering at Georgia Tech said in a statement.
The system they came up with is based on a PR2
mobile manipulator originally manufactured by Willow
Garage. It’s a wheeled robot with two arms and a
head, and it can manipulate objects like water bottles,
washcloths, hairbrushes, or an electric shaver.
The robot can be controlled using assistance
technologies that users are already familiar with, such
as eye tracking or head tracking. The interface shows
a “robot’s eye view” from cameras inside the robot’s
head, making it easier for the user to control.

24 SERVO 03/04.2019
Poetry in Motion

T his pneumatic, lightweight robot


arm from Festo can be used for
numerous applications, with either
free and flexible movements or
defined sequences. In combination
with various adaptive grippers, it can
pick up and handle a wide variety of
objects and shapes. At the same time,
it’s completely compliant and poses
no danger to the user — even in the
event of a collision.
The BionicSoftArm owes its
flexibility to its modular design,
which can be combined from several
pneumatic bellows segments and
rotary drives. Depending on the
requirements, the length of the
BionicSoftArm can be varied with up to seven pneumatic are implemented by a Festo Motion Terminal VTEM, which
actuators, thus providing maximum flexibility in working range makes the control and regulation of complex kinematics
and mobility. This makes it very easy to implement applications possible in the first place. Through the internal control
that are difficult to realize with a standard robot. algorithms of its motion apps and the installed piezo valves,
This allows the BionicSoftArm to work around flow rates and pressures can be exactly dosed and varied to
obstacles even in the tightest of spaces. Direct human–robot any setting in several channels simultaneously. This enables
collaboration is just as possible as use in classic SCARA both powerful and fast as well as soft and sensitive motion
applications; for example, pick-and-place tasks. The elimination sequences.
of costly safety devices such as cages and light barriers The interface between the tablet and the Festo Motion
shortens conversion times and thus enables flexible use at Terminal is the ROS (Robot Operating System) open source
different locations — completely in accordance with adaptive platform on which the kinematic’s path planning is calculated.
and economical production. In addition, ROS interprets the incoming code from the tablet
The bellows are made of sturdy elastomer. Each one is and forwards the resulting axis coordinates to the Motion
covered with a special 3D textile knitted fabric which consists Terminal.
of two layers. A soft knitted fabric lies directly on the bellows
to protect them
from friction and
wear. The high-
strength fibers are
oriented so they
allow the bellows
to expand in the
required direction
of movement and
at the same time
limit this in other
directions. Thanks
to this new fiber
technology, the
power potential
of the entire
kinematics can be
exploited.
The commands

SERVO 03/04.2019 25
We Do Windows

M itsubishi Estate Co., Ltd., in collaboration with


Japan Biso Co., Ltd., Shin at the Marunouchi
Building, recently conducted a demonstration of an
outside window cleaning system utilizing an articulated
arm equipped with a “multi-purpose wall work robot.”
The size is 3,200 x 940 x 2,000 mm (width x
depth x height). The inside gondola is equipped with a
multi-joint robot arm with seven degrees of freedom
to provide a wider range of movement. It is made by
Yaskawa.
The body is composed of the base system and the
working head; the travel system provides further lateral
movement.

Fly, MetaFly, Fly

A French inventor generated quite a bit of buzz when he


took to Kickstarter recently to raise funds for an insect-
inspired winged drone called MetaFly. At the time this was
reported, more than 1,850 people had pledged more than
$187,500 to bring the drone to market.
Unlike traditional commercial drones which use
propellers to generate lift, MetaFly uses its wings to take
flight. Much like the bees it’s modeled after, MetaFly flaps its
wings vigorously, creating a differential between draft and lift
— an efficient flight mechanic used by flying animals.
The drone is lightweight and maneuverable, and can
fly up to eight minutes on a 12 minute charge. MetaFly can
reach speeds up to 11 mph. The remote has a wireless range of 320 feet.

26 SERVO 03/04.2019
Internet from the Sky

A n experimental solar-powered drone capable of beaming Internet down to earth is set to take off.
The drone — called the Hawk 30 — is the product of a tightlipped partnership between Softbank
and AeroVironment, an American aerospace company. The joint venture is aimed at delivering 5G and IoT
(Internet of Things) connectivity from the skies.

Get Racey

D rone racing has been a thing for a


few years now, and like any new
sport, it’s continuing to look for ways
to grow its audience.
Following earlier deals with
the likes of ESPN and Sky, the sport
appears to be turning away from pay-
TV channels to offer the fun for free.
From August 2019, long-time fans
and newbies alike can hit Twitter to
enjoy all the action from the fourth
Drone Racing League (DRL) season,
billed as the world’s only professional
drone racing circuit.
Along with NBC and NBCSN, Twitter will be streaming the 2019 DRL
Allianz World Championship Season, premiering on Sunday, August 11 at 2
p.m. ET.

SERVO 03/04.2019 27
Help Me! I’m Melting!

R oboticists at Colorado State


University have developed a small
walking robot that can melt and solidify
its bones on-the-fly to optimize its legs
for different motions.
Jianguo Zhao, who worked on
jumping robots as a graduate student
at Michigan State, is now an assistant
professor at Colorado State. In a paper
published in IEEE Robotics & Automation
Letters, Zhao introduced a new kind of
small reconfigurable robot that relies on
a plastic structure that can be selectively
melted and re-hardened to change joint
configurations and the resulting motion.
The researchers call these “shape
morphing joints.”
It takes just over 10 seconds to
soften the PLA that the robot’s joints
are made of using a wrap of wire that
heats up when voltage is applied to it.
A silicone cover helps the PLA hold its shape while it’s compliant. Depending on which of its joints are softened, the robot can
switch between a rigid structure, a four-bar linkage, and a five-bar linkage. Cool thing is, you get all this with just two joints, and
it’s easy to add more.
This particular robot is about as simple as a walking robot can get, while still being able to take advantage of these new
joints. It’s mostly 3D printed and only uses two actuators in total to move all four legs.

28 SERVO 03/04.2019
Man’s Best Friend
A company called Tombot has come up
with a way to help improve the quality
of life for seniors facing depression and/
or loneliness: a robotic companion dog
(called the Tombot Puppy) that behaves and
responds like a real pup, but without all
the responsibilities of maintaining a living,
breathing animal. The company even enlisted
the talented folks at the Jim Henson’s
Creature Shop to help make the robo-dog
look as lifelike as possible.

Make Some Noise neural network ingested the

A t Georgia Tech’s Center


for Music Technology
(GTCMT), Gil Weinberg
following:
• 10,000 files from 15
improvisational musicians
and his students have a lot playing responses to
of experience with robots different emotional queues.
that make noise of various • 300,000 samples of
sorts. They’ve now used a musical instruments playing
new deep learning-based different musical notes, to
technique to teach their add musical expressivity to
musical robot, Shimi a basic the spoken word.
understanding of human • One of the rarest
emotions, and how to languages in existence: a
communicate back in just the nearly extinct Australian
right way to those humans using music. aboriginal vernacular made up of 28 phonemes.
Shimi has been around since 2012, and initially, it was
programmed to play music that it would analyze to be able Hmmm, that last thing seems a little bit odd, doesn’t
to dance along to, utilizing its few (but effective) degrees of it? Well, it’s key to Shimi’s vocalizations because it’s how
freedom. the robot makes sure that nobody (or, almost nobody) can
Training Shimi to be able to improvise effectively in ascribe any direct meaning at all to the noises that it’s making.
voice, tone, and motion has involved training it with lots and
lots (and lots) of data. To develop Shimi’s behaviors, a deep

SERVO 03/04.2019 29
Exercise Coach, Robot-Style
H ere’s a project to develop a robot coach capable of
demonstrating rehabilitation exercises to patients. It watches
patients carry out the exercises, and then gives feedback to improve
performance and offer encouragement.
Poppy is an open source and anthropomorphic robot, which
has a sufficient number of degrees of freedom to perform coaching
exercises. The goal is to increase the time patients spend exercising,
by alleviating the lack of time a physiotherapist can spend monitoring
a patient.
The KERAAL project is a consortium composed of three
partners: Telecom Bretagne, Generation Robots, and CHRU Brest.

Locomotion Notion

U sing the fossil and fossilized footprints of a


300-million-year-old animal, scientists from
EPFL and Humboldt-Universität zu Berlin have
identified the most likely gaits of extinct animals
and designed a robot that can recreate an extinct
animal’s walk.
The team built OroBot, modeled off living
reptiles and amphibians, to examine how animals
that predated the dinosaurs evolved to walk on dry
land.
This study can help researchers better
understand how vertebrate locomotion evolved
over time.

30 SERVO 03/04.2019
Kroger Really Delivers
T he Kroger Co. (one of the largest
grocery retailers) and Nuro, a robotics
company dedicated to transforming local
commerce, announced the recent launch
of their autonomous grocery delivery
service in Houston, TX. The innovative
service will be available through two
Houston Kroger stores, reaching
customers residing in four local zip codes.
Since August 2018, Kroger and
Nuro have operated a first-of-its-kind
self-driving grocery delivery service in
Scottsdale, AZ, servicing a single zip code
with an autonomous vehicle fleet. In
late 2018, the partnership expanded to include Nuro’s custom vehicle, R1, marking the nation’s first fully
unmanned delivery service available to the general public.
Together through this pilot program, Kroger and Nuro successfully and safely completed thousands of
deliveries to customers in Scottsdale. With the new launch, Kroger and Nuro will transfer the autonomous
grocery delivery program to Houston for the next phase of the pilot.
The Houston program mirrors last year’s pilot in Scottsdale in that Nuro has started off by using
its autonomous Toyota Prius vehicles. Later this year, it will switch over to its much cooler “R1” delivery
pod. Both vehicles come with all the usual gizmos — sensors, cameras, GPS, and so on — to ensure their
safe operation as they navigate Houston’s public roads. The R1 is about two-thirds the length of a Toyota
Corolla and has a top speed of 25 mph.
Kroger customers can use an app to place orders seven days a week, with deliveries able to be
scheduled for the same or following day. Delivery costs $5.95 with no minimum order required.
Nuro co-founder Dave Ferguson said customers’ response to its self-driving delivery vehicles in
Scottsdale had been “enthusiastic,” adding that his team is looking forward to showing off the technology
to shoppers in Houston in the coming months.

A No-Brainer
B rain Corp, a San Diego, CA
based provider of autonomous
navigation systems, has announced
a partnership with Nilfisk: a
Copenhagen, Denmark-based
cleaning solutions provider that has
been around for 110+ years.
Nilfisk will license the BrainOS
AI platform for the production,
deployment, and support of its
robotic floor cleaners. According to
a Brain representative, “There are no
Brain-Nilfisk products on the market
currently.” Nilfisk is developing the
Liberty SC50 autonomous scrubber
and will add Brain’s technology and
expertise to the mix going forward.
Nilfisk Liberty SC50 autonomous scrubber. Image courtesy of Nilfisk.

SERVO 03/04.2019 31
Digits Deliver
D igit — the humanoid creation of Oregon-based startup, Agility
Robotics — boasts a bipedal design (meaning that it walks
on two legs) that makes it capable of carrying boxes weighing up
to 40 pounds, pushing open doors, and generally maneuvering in
environments like a real flesh-and-blood delivery person.
“Only a very small percentage of residential properties in the
US are wheelchair accessible,” Dr. Damion Shelton, CEO and co-
founder of Agility Robotics, told Digital Trends recently. “Assuming
that this is a reasonable proxy for accessibility by wheeled ground
vehicles in general, this means that the overwhelming majority of
homes are not accessible to wheeled delivery robots. The one
commonality that we know about all delivery locations today
is that they are — almost without exception — accessible to
pedestrians.”
Despite the talk of being ready for the market, Digit isn’t
available to purchase just yet. The version currently being shown
is only version 1.0. The next revision is currently underway, and
will make its grand debut later this year along with an indication
of pricing.
“We’ve announced that we anticipate the first deliveries to
customers occurring in Q1 2020, and don’t have anything to add
to that timeline,” Shelton said.

Artist’s rendering
of the Europa
“tunnelbot.”
Image courtesy of
Alexander Pawlusik,
LERCIP Internship
Program NASA
Glenn Research
Center.

To the (Jupiter’s) Moon

A group of scientists wants to send a nuclear-powered “tunnelbot” to Europa to


blaze a path through the Jovian moon’s thick shell of ice and search for life.
Europa, the fourth largest of Jupiter’s 53 moons, is supposed to be one of the
best candidates in our solar system for hosting alien life. Researchers believe that its
icy crust hides a liquid water ocean and that vents through that crust might deliver
the necessary heat and chemical ingredients for life into that ocean.To peek beneath
the thick veil of ice, researchers on the NASA Glenn Research COMPASS team
(a group of scientists and engineers scattered around the country and tasked with
solving problems for NASA) think they have come up with a tunnelbot that will work.

32 SERVO 03/04.2019
Can You Feel It?

R esearchers from EPFL, the


Sant’Anna School of Advanced
Studies in Pisa, and the A. Gemelli
University Polyclinic in Rome have
developed a next-generation bionic
hand that allows amputees to regain
their proprioception (perception
or awareness of the position and
movement of the body). The results of
the study (which have been published
in Science Robotics) are the culmination
of 10 years of robotics research.
The new device allows patients
to reach out for an object on a table
and to ascertain an item’s consistency,
shape, position, and size without having
to look at it. The prosthesis has been
successfully tested on several patients
and works by stimulating the nerves in
the amputee’s stump. The nerves can then provide sensory feedback to the patients
in real time, almost like they do in a natural hand.
Current myoelectric prostheses allow amputees to regain voluntary motor
control of their artificial limb by exploiting residual muscle function in the forearm.
However, the lack of any sensory feedback means that patients have to rely heavily
on visual cues.

Learning to Adapt

A fter Japan’s Fukushima nuclear disaster


in 2011, robots were sent into the
power plant to clear up radioactive debris in
conditions far too dangerous for humans. The
problem, says robotics researcher Professor
Jean-Baptiste Mouret, is that the robots kept
breaking down or came across hazards that
stopped them in their tracks.
As part of the ResiBots initiative, he is
designing a lower-cost robot that can last
long periods without needing constant human
maintenance for breakages and are better at
overcoming unexpected obstacles.
The ResiBots team is using what it refers
to as micro-data learning algorithms which can
help robots adapt in front of one’s eyes in a
In trials, the ResiBot robot learned to similar way to how animals react to problems. An animal will, for example, often find a
walk again in less than two minutes way to continue moving if they get injured — even if they don’t know exactly what the
after one of its legs was removed. problem is.
Photo courtesy of Antoine Cully/
In contrast, most current robots self-diagnose a problem before working out a
Sorbonne University.
way to overcome it, says Prof. Mouret, principal investigator at ResiBots and a senior
researcher at the Inria research centre in France.
‘We’re trying to shortcut this by finding a way for them to react without
necessarily having developed an understanding of what’s wrong,’ he explained.

SERVO 03/04.2019 33
Vortex – The Shell
Spinner Adventure By AJ McNamara

I
f you’ve ever watched
BattleBots™, you’ve probably
seen what people call a shell Vortex Mk.1 base.
spinner, or full body spinner. The
robot’s exteriors are large metal
shells that spin at incredible speeds.
They vaguely resemble a Roomba
that’s been turned into a machine
of destruction. Ever since I first saw
BattleBots, shell spinners caught my
eye. Not only did they look cool, but
they could deliver some of the best
hits I’d ever seen in the ring.
I’ve always been interested in
robotics and had competed in various
competitions throughout grade school.
Back in 2015, I learned that smaller
scale robot competitions existed not
too far from where I lived. A group
of my friends and I set out to build a
combat robot for the Central Illinois We went winless at the Since I didn’t have much access to
Robot Brawl in the summer of 2016. competition, but still had a ton of fun. machining and tools at the time, it was
Underestimating the difficulty of this I was instantly hooked on building designed around using mostly hand
task and overestimating our abilities, combat robots, and I determined I tools and a few parts from a friend
we did most of the building in the two wanted to build a shell spinner like the of a friend who had access to more
weeks leading up to the tournament robots I had admired on TV. advanced tooling.
and weren’t finished till the day of. My planning began that fall, but The robot had an entirely
I didn’t really know where to start. aluminum structure and used a steel
The concept itself was simple: Make cake pan as the shell. Vortex was

FEATURED the outside of the robot spin while


the insides stay put. As I soon learned,
constructed in the summer and I
planned to compete at CIRC 2017.
that task was far easier said than With the robot fully assembled, I
THIS MONTH done.
Being the poor student that I was,
wanted to test it out. I turned the
controller and the bot on, and started
I settled on building a Beetleweight (3 to drive forward.
34 Vortex — The Shell lb) robot. It would be small enough for The robot drove around a bit
the cost to be manageable, but large before my ears were met with the
enough to not making manufacturing sickening sound of grinding gears.
Spinner Adventure
and assembly a nightmare. I settled Some screws were incorrectly mounted
on the name “Vortex” for the robot in the drive gear boxes and the gears
38 The Bot of Theseus because of its spinning nature, plus, had stripped. The competition was
it was the name of my high school only a day away, and I didn’t have
Or “The Sparadox” robotics team. enough spare parts to fix the robot in
Over the next few months, I had time.
gathered a list of what parts I thought It was a disappointing start to
42 I Build BattleBots I needed and began to purchase them. my robot, but I was determined to

34 SERVO 03/04.2019
To post comments on this article
and find any associated files
and/or downloads, go to www.
servomagazine.com/magazine/
issue/2019/02.

Vortex Mk.2 internals. Vortex Mk.2 with shell.

as a shell spinner speed above that limit; it was plenty


succeed! I decided it was time to make has considerable rotational inertia, powerful enough as-is.
some major changes to the design. which is how shell spinners store so That left only one glaring issue:
This year, I had access to my college much energy. Getting the moment the batteries. The batteries were one
machine shop — a huge resource in of inertia (MOI) to be perfectly in the of the few components that I held
my bot-building journey. Aside from center is a hard task. Since the shell over from the original robot and
the weapon motor, drive ESCs, and was made from so many individual retaining them turned out to be a
battery, everything was new. The pieces, it took a lot of testing and bigger issue than expected.
heavy metal baseplate got swapped adjustments to get the shell to be Originally held in by Velcro™
for a custom-cut piece of fiberglass. stable when spinning. straps, I soon learned that was far too
The parts that were previously cut Most of the stabilization was weak of a solution. In testing, one
with a Dremel and drilled with a hand done by drilling holes in the side of battery had its connector plug stripped
drill were now done with a waterjet the shell to remove material where it off the wires by the shell, and another
and drill press. The cake pan shells was too heavy. This did leave some had nearly been sliced in half by the
were replaced with welded aluminum un-cosmetically pleasing holes, but the shell when the Velcro loosened. The
shells that sported AR400 steel teeth. robot no longer wobbled around as battery was too thick to have a rigid
The robot was finally something I was long as the throttle was below 60%. cover placed over it, and no retaining
proud of; it was a miniature version of I decided not to turn the weapon mechanism we had on hand would
the bots I had seen on TV. keep the batteries and all
Vortex — now dubbed the wires contained. Since
version “Mk.2” — was Balancing holes in shell. the top of the robot was the
slated to compete at HORD shell (which had holes to
2018 in June, but in the save weight), anything that
weeks leading up to the came loose inside the robot
tournament, new-found risked being chewed apart by
problems arose. the robot’s own weapon.
A few initial tests of At the tournament, the
the weapon in an impact robot ate one more battery
resistant test box helped dial in testing. Since the robot
in the weapon belt tension, had a nasty habit of being
but two major problems close to a LiPo battery fire
persisted. The first was the every time it was turned on, I
balancing of the shells. A decided not to compete. The
weapon as large and heavy robot was an actual safety

SERVO 03/04.2019 35
that Vortex could take hits from other
Beetleweight rumble at HORD. robots and stay on its wheels.
The next competition was CIRC
2018; the tournament that started
it all two years prior. However, a few
upgrades were planned before then.
The bearings were upgraded for more
stability, and new batteries with a
3D printed plastic clip were added.
Perhaps most important though,
Vortex got a new paint job. Black
with electric green stripes to make the
robot look sleek.
Vortex arrived at CIRC fully
functional and ready to rock for
the first time. In its first real match,
Vortex faced off against Tidal Wave:
a blue drum spinner with two large
wheels. Vortex delivered some great
hits. It ripped the drum off the
opponent robot and knocked it out
in only 26 seconds. After two years
of development, Vortex won its first
match. I was through the roof with
excitement.
Hot on the heels of this success, I
prepared to face off against my next
opponent, Ele-Flip: a robot with a plow
for lifting opponent robots. The first
Vortex vs.
Tidal Wave.
few hits knocked Ele-Flip around quite
a bit and pulled the plow loose from
the robot. In just under a minute, Ele-
hazard. HORD was still fun, and I hung for the duration of the match. When Flip was immobilized, and Vortex had
out with my team as they competed. the five minutes were up, Vortex was its second knockout of the day.
When the time for the rumble one of two robots still moving. The The next match, however, would
rolled around, I decided to give Vortex crowd voted for the other robot to win break Vortex’s streak. Vortex faced off
one last shot. I used every method because he pushed more opponents against Sparrow: a D2 wedge robot.
available to secure the battery inside into the pit, but I still considered this a Vortex was knocked into the pit after
the robot: epoxy, superglue, gorilla great success for the robot. It showed getting caught on Sparrow’s wedge. A
tape, and steel zip ties disappointing way to go
(thanks to the Russian out, but Mario from Team
Winter team). I was Sparrow had played very
approved to enter the well.
Beetleweight rumble, Finding myself in the
which is a giant free- loser’s bracket, it was win
for-all at the end of the or be eliminated at this
competition. point. My next opponent
After delivering a few was Chunky Monkey:
good hits in the rumble, another D2 wedge kit.
one of the weapon The match started alright,
motor wires shorted and Vortex did get flipped
out. With the drive still upside-down for the first
functional, I was able to time, but the self-righting
push and take damage Painted shells for CIRC. arm worked perfectly,

36 SERVO 03/04.2019
and Vortex got back on its
wheels. Ultimately, however,
Vortex faced the same fate as
before. With a few seconds to
go in the match, Vortex was
knocked into the pit, giving the
win to Chunky Monkey.
It was a bit disappointing,
but going 2-2 at its first
competition wasn’t horrible.
Plus, the rumble was still to
come. The rumble was a chaotic Beetleweight
four minutes; Vortex’s shell died rumble at CIRC.
in the first 20 seconds, and at
some point, the top of the shell
and the ring came completely
apart, but the bot battled on.
After a few robots got knocked
into the pit and several of them
lost mobility, there were only a
upgrades and get back in the ring! I
liked shell spinners when I first saw
them on TV many years ago, and after
building my own, I love them. SV

Vortex vs. Ele-Flip.

few robots still able to move. win the rumble, with Vortex’s shell
With two minutes left, only Vortex taking blow after blow from Ataxia’s
and a horizontal spinner named Ataxia blade. With 11 seconds remaining,
were alive. The two robots battled to Ataxia bounced off Vortex and flew
into the pit. Vortex
had knocked
Vortex after the rumble. out at least four
robots, but more
importantly, it had
survived. My shell
spinner that had
been such a long
series of trials and
struggles had won
a rumble.
It’s been a
long journey to
get a shell spinner
working, and
I’ve learned a lot
along the way. I
can hardly wait
to make more

SERVO 03/04.2019 37
The Bot of Theseus
Or “The Sparadox” By Stephen Pasek

I
n combat robotics, damage to identical bots between fights. This The “Replace” School
your robot is expected; for a bot article attempts to look at both sides The opposing philosophy is
to leave the match unscathed of this argument and discuss the the “replace” way of doing things.
is highly unusual. Aesthetic merits of both philosophies. Builders in this school of thought
damage to the bot such as scuff think of tournaments as a testing
marks on armor or a slightly bent ground for their designs. They aim
frame is little cause for concern. It’s a
Schools of Thought to test their design in peak condition
much different story when the delicate The “Repair” School against different designs also in peak
internals of a robot are damaged. The first combat robot philosophy condition to determine which models
Back in the day when BattleBots™ we’ll be looking at is what I’ll call are superior. For these builders, all the
were made out of sticks, stones, the “repair” philosophy. This school work is done prior to the competition
and sheet metal, almost anything on of thought views combat robot in trying to optimize a design to be
the bot could be repaired with duct tournaments as a competition focused successful against the widest range
tape and wires. Presently — with the on endurance and durability. The bots of opponents. Once satisfied with a
adoption of modern electronics, such are designed to be simple, tough, design, they make multiple copies of it
as brushless motors, and advanced and easy to repair. Students of this and bring those to competitions and
manufacturing techniques like 3D school view tournaments as a test of use the time in between fights to relax
printing and CNC machining — the builder’s skill in repairing their bot and think of improvements to their
repairing a bot is much more involved. under pressure and their versatility in design.
It’s no longer feasible to repair a coming up with ad hoc fixes. These builders view combat
brushless motor or its ESC (electronic One way to determine a ruling robotics more as an engineering
speed controller). on spares is to look at other similar challenge like auto racing. Teams
Replacement parts are most sports. Followers of the repair school design and build a vehicle to race and
builder’s only option. With the compare combat robotics to fighting have multiple copies and spares in
growing popularity of CNC machined sports, like boxing or MMA. There are case anything goes wrong. The focus
chassis and 3D printed unibodies, two opponents in a ring and they try is about the design and building skills
repairing a robot’s frame while at to knock the other out through direct of the team and trying original ideas;
competition is almost impossible. damage. If combat robotics is like not in the ability of the team to repair
The most recent development has boxing, then swapping out a damaged the car during the race. However, race
been builders bringing multiple copies bot for an identical one between fights cars aren’t supposed to crash, so they
of their bots to a competition and is the same as a boxer’s twin going in aren’t designed with repairs as such
swapping them out when one gets too for him/her in the next round when a high priority. If you crash in a race,
damaged. Examples of this are Bronco the first twin gets knocked out. This is you’re out; whereas if you get hit in
in BattleBots Season 2 (2016), and a clearly not the case in boxing. combat robotics, that just means the
Facebook post by Valkyrie (October A fight in boxing is meant to test match has begun.
2018) showing two of each Antweight the endurance of the fighter and how
and Beetleweight bot they were many hits they can take versus how
bringing to the Franklin Institute fights. quickly and devastatingly they can dish
Potential Rule Ideas
There was a large response from them out. Boxers don’t usually get Almost all builders agree that
the community at this post, with many ripped in half during a fight though, banning spare parts in their entirety is
builders viewing this as being against and have several weeks or months foolish. Most of the objection comes
the philosophy of the sport. They between fights to heal and recuperate; from the number of spares brought
argued that while at a tournament, unlike BattleBots, where you can have and the degree to which they are pre-
builders should try to repair their bots as little as 30 minutes between fights assembled. Practitioners of the replace
as much as possible and not substitute for repairs. school obviously don’t want any rules

38 SERVO 03/04.2019
limiting the number and Speaking of the rushed
arrangement of spares they nature of most combat
can bring to matches, so this robotics tournaments,
will focus on what ideas those there’s a clear benefit to
of the repair school have teams having an extra fully
considered in order to level functional bot. If their primary
the playing field. The objective bot is heavily damaged —
is to allow spares such that no but repairable — instead
team has an unfair advantage of trying for hours to fix it
by being too well-equipped. before their next fight, they
Some methods considered are can just use their secondary
to limit spares by weight, cost, bot immediately. This allows
number, or type. fights to proceed much
Limiting spares by weight quicker. Spare bots mean
is relatively straightforward. faster turnarounds. Faster
Each robot would be allotted turnarounds mean more
a certain percentage of its fights. More fights mean
weight class for spare parts of more fun.
any kind. For example, with
a 50% allotment for spares,
a 30 lb vertical spinner could
What Makes a
bring up to 15 lbs of spare Double Foils. Bot “Itself”
parts. These 15 lbs could be
divided amongst a 5 lb disk, two 3 more work to the already full plate If you replace a broken
lb motors, and a spare armor piece of the EOs. If they have to weigh or component in a bot, is it still the same
weighing 4 lbs. count up spare parts, it would take bot? If you replace all the motors in
Limiting spares by cost is many times longer to go through your bot, to the outside observer it’s
potentially the most difficult method to safety inspection and the number of still the same robot. If you swap out
enforce. How I imagine it will be set up fights held would drop. The other a disk weapon for a bar, most people
is that the Event Organizers (EOs) will method — setting a budget for each would agree that it’s the same bot,
announce a total cost per bot when team for an event, and allowing them just with a different weapon. But
the registration for the fight opens. to either spend all that budget on one what if you start swapping out frame
Upon arrival to the event, builders bots worth of parts or make a cheaper components or changing the weapon
will have to provide receipts proving bot and bring lots of spares — has in a significant way? Is it still the same
they did not go over the agreed upon plenty of issues as well. bot then?
amount. If they went over, spare parts How do you ensure the teams This is a question both sides of
will be confiscated one-by-one until the followed the budget set forth? What is the argument have for the other.
price limit is reached. included in the budget? Are machining Those on the replace side think
The last methods will probably be and manufacturing costs included? that repairing a bot and changing
combined. Builders would be allowed Having your own mill or lathe or it significantly between fights is, in
a certain number of each type of spare having access to a machine shop for fact, using a different bot. Those on
components; e.g., four spare drive cheap offsets a large portion of the the repair side think that replacing
motors, one spare weapon motor, two cost of the bot. This would give an damaged components on your bot
spare motor controllers, one spare advantage to more established teams with undamaged ones makes your bot
receiver, etc. This would be relatively with more resources and hurt teams not the original.
easy to enforce provided that EOs are that are just starting out. This is an issue in every weight
permitted to examine all the toolboxes What sort of advantages for class. However, it’s most visible on the
and containers the teams bring in. teams can be quantified in dollars and televised shows like BattleBots and
regulated to level the playing field? Robot Wars. In BattleBots Season 1 on
What about education? A five-person ABC, Biteforce was primarily a lifter/
Practical Limitations team of engineers with four year grappler bot, but in the final match
to the Rules degrees has literally a million dollars of against Tombstone, they added a large
knowledge. At this point, there is no steel wedge to their rear end and used
Any addition to the rules is adding way to balance fights based on cost. it to deflect Tombstone’s blade while

SERVO 03/04.2019 39
keeping their lifting arms unused. conventional BattleBot is designed On determining if a bot has been
Was this still the same Biteforce around one main weapon type, e.g., repaired or changed beyond
that fought previously? In Season 2, Tombstone is a horizontal bar spinner; recognition:
Biteforce returned with an entirely new Nightmare is a vertical disk spinner; I liked the suggestion put forward
design. Instead of a lifter/grappler, etc. by Hal Rucker, an experienced builder
it was now a vertical spinner. The A Swiss Army bot, on the who competed on BattleBots Season
announcers of the show treated this other hand, is designed to have 2 with Ringmaster and Season 3 with
Biteforce as the same one that won interchangeable weapon types. DUCK!. His solution was to have a
the previous season, but clearly this Bombshell in Season 2 of panel of five seven-year-old kids look at
was not the same bot, though the BattleBots is a prime example of a picture of the bot taken when it was
paint scheme and name were the a Swiss Army bot. They had four registered and compare it to the bot
same. different weapon configurations: about to enter the match.
Tombstone, on the other hand, lifter; axe; horizontal bar spinner; and If the kids agree by majority vote
has remained mostly unchanged vertical disk spinner. that the robot in front of them is the
externally forever. Other examples of They ended up using three of same one that’s in the photo, the
teams adding things onto their bot those weapon types over the course bot gets to go into battle. The only
before fighting a specific opponent of the show to great effect, coming problem with this idea is if there is
are: in second place. In using different a shortage of seven-year-olds at the
primary weapons, is Bombshell a competition. This rule might mean
• Ghost Raptor adding the De- different robot each time? that Bombshell in Season 2 could
Icer fork to keep themselves I would argue that it is not, since not compete in its different weapon
out of reach of Icewave’s it was designed to use those different configurations.
spinning bar. weapons and therefore had to make
• Bronco adding a flat extension some sacrifices to be more flexible. On whether or not rules are
to the end of their flipper arm What if you had a robot like Bombshell needed for this issue:
to reach under Tombstone’s but just added a drive chassis to all the Paul Ventimiglia, builder of
weapon bar and flip him. weapon modules, effectively making Biteforce, has this to say: “... what
• Carbide adding a metal chain four separate bots. I would argue that others do doesn’t affect you. Shut
guard after their first fight these are separate robots and could up and build and compete however
against Nuts to prevent Nuts’ not compete under one name. The you want. Other people building stuff
flails from breaking their chain success of one would not mean a doesn’t make you lose or make you
again. success for all. upset. You make yourself upset and
• Tombstone itself also has make yourself lose.”
different bar geometries to I like his message, which I take to
use against different types of
The Pros Weigh In mean as other team’s preparedness is
opponents. This is not the first time this topic not a reason for them to be restricted
has been discussed by the community, to cover your unpreparedness.
Season 2 of BattleBots also and several well-known builders have There will never be a consensus on
showcased a previously unseen style shared their opinions. what is fair, and trying to restrict the
of robot: the Swiss Army bot. A amount or configuration of spare parts
teams bring will not work to make
things better, and will stifle creativity in
the sport.

Someone of the “repair” mindset:


Rory Mangles, builder of Nuts
for Robot Wars, has truly seen what
can be accomplished on a shoestring
budget. In previous seasons of Robot
Wars, Nuts was seen as a joke bot,
and fared poorly in matches. This
all changed in its third appearance
on the show, when it showed off its
Extra bot made from spare parts. new Melty Brain control system and

40 SERVO 03/04.2019
proceeded to defeat the previous competition, but we were able to get other for three minutes?
champion, Carbide, in one of the it functional enough to pass safety If I had an infinite budget to
greatest upsets in Robot Wars’ history. within a few hours of arrival. go on BattleBots, I would definitely
This is his view on this debate: “Getting Unfortunately, the design of the bring as many identical copies of the
a ruined bot running again in a tight bot was very low on space and the same bot to the event as I could. You
time frame and with limited resources frame was machined from a single start designing your robot with no
is by far my favorite part of the sport ... billet of aluminum, so our ability to specific enemies to fight, so you try
But then, I’m weird like that.” repair it at the event was severely to make your design work against all
limited. We had to forfeit our fights for weapon types. Then, you take it to
the weekend. If we could have gotten a competition and see how it fares
My Thoughts it running before the competition, against each bot type.
I was a participant in Season 2 of we would have needed to bring the For other builders, the goal of
BattleBots on a team with some of my equivalent of an entire extra bots’ the competition is in maintaining a
classmates from college. It was our worth of spares in case of complete successful bot and testing the builder’s
first time building any sort of combat destruction. So, I can also support the resourcefulness and technical skill
robot and we jumped right into the replace school of thought. under pressure. Should those that
deep end. What is the end goal of a enjoy repairing their bots be the only
We never ended up fighting the tournament? Is it to see which bot is ones whose voices matter? I think
bot we brought since we spent the the toughest and most deadly? Or, is it doesn’t matter as long as you’re
entire time trying to get our bot to it to test your design and building skills having fun.
work in the first place. against dozens of different designs?
Even so, it was still one of the Should there be an additional two
most fun experiences I have had classes: Fun and Competitive? The Fun
Conclusion
and it made me very interested in class would allow multiple copies of Is this discussion even necessary?
combat robotics in general. I totally the same bot while the Competitive The ultimate goal of combat robotics
understand the viewpoint of those class would only allow repairs and no should be to have fun and learn new
builders who prefer repairing their bot spare parts. things.
at competition. The race against the BattleBots and Robot Wars are Whether having fun to you is
clock to get your bot up and running is shows; they are broadcast to a large making a bot that’s indestructible or
an amazing thrill. audience and therefore need to be making a bot that tries new design
A few years later, I made a 30 entertaining. Is it more entertaining to methodologies and testing them under
lb Melty Brain bot to compete at see two bots at the top of their game different scenarios, the rules shouldn’t
Motorama. Again, the bot wasn’t go head-to-head or to see two barely- stifle creativity or favor one design over
complete when we arrived at the held-together bots limp around each another. SV

EASILY ADD MOTION TO


YOUR PROJECT WITH THE
ACTUONIX IOT WIFI BUNDLE

ACTUONIX
R SERIES IOT WIFI WIFI
MICRO LINEAR ENABLED
ACTUATOR BUNDLE DEVICE

VISIT ACTUONIX.COM FOR PRODUCT


INFORMATION AND PROJECT RESOURCES

SERVO 03/04.2019 41
I Build BattleBots By Andrea Gellatly (Suarez)

I
’m an engineer and subtler than anyone imagines.
robot builder. You most Whenever I introduce
likely didn’t hesitate to myself as an engineer, I
believe that statement, often get a surprised look
because you can’t see and a “Really?” that I have
that I also happen to be a to answer with evidence of
woman. My name is Andrea my merit as an engineer.
and I am captain of Team The comment section of our
Witch Doctor on the TV BattleBots posts are often
show, BattleBots™, where sprinkled with doubt about
we compete our 250 lb robot whether my gender is the
against the best in the world. reason we were selected for
I’ve received quite a the show, favored in a judges’
bit of attention as a female decision, etc. Every puzzled
captain. Most of it has been look instills doubt in my
overwhelmingly positive and own abilities and is another
encouraging. I’ve had the reminder that I’m doing
privilege to speak to all kinds something that isn’t expected
of BattleBots fans from all of me.
walks of life. I always get Because of this reality,
the same question: “How I’ve made a very conscious
did you decide to become effort on BattleBots to
an engineer?” As a woman, “show” instead of “tell.” I
my answer is expected to avoid every interview question
hold some unique wisdom about being a female robot
or untapped secret. I’ve Team Witch Doctor, as seen on BattleBots on Discovery builder, and refocus on the
never had a good response channel. Photo by Daniel Longmire. robot and the tournament.
because I don’t have a There’s no better way to help
singular moment when I change perceptions than by
suddenly decided to be an doing what I do best: Building
engineer. I’ve been working robots. Besides, young girls
on answering this question for watching the show don’t
a long time. need to be told what is right
No one has ever thrown in front of them: If I can build
their arms in the air and 250 lb combat robots, so can
yelled that I don’t belong. they.
I often reassure young girls High school is a defining
who are anxious about time for most kids, and the
entering the STEM (Science, same was true for me. I
Technology, Engineering, and struggled to choose a school
Math) community that my and went to a slew of open
daily experience as a woman houses trying to find the right
in STEM has truly been fit. I found my answer at an
wonderful. BattleBots fans around the world were shocked to see a unexpected place: Carrollton
The hardest part of being brightly colored team with a female captain make it to the School of the Sacred Heart, a
a woman in a STEM field is quarterfinals in Season 1. small all-girls catholic school.

42 SERVO 03/04.2019
Team Mean
n’ Green won
second place at
BattleBots IQ,
with our full-
body spinner
Marvinator, as
high school
seniors in 2006.
This remains
a record for
an all-female
BattleBots
team!

hadn’t seen the majority of


the school, but I knew that
this was where I belonged.
By the time it became
obvious that an all-girl
robotics team was unusual,
I was too committed to
doubt whether I belonged
in this sport. It was
empowering to see that
My first fighting robot, 120 lb Kerminator,
as a high school sophomore in 2004.
I could imagine a design
We had to learn how to weld, machine, and then make it happen.
cut, grind, and much more to build this The more I learned — CAD
machine! It didn’t win any awards, but it design, machining, welding
proved to us that we were robot builders. — the more power I had
Photo by Daniel Longmire. to make better robots, so I
decided to study biomedical
The open house was run mostly by engineering. After all, the
the students, with groups of girls human body is just another
exhibiting their classes and clubs. machine; an incredibly complex
While walking down a hall, I heard robot. I like to joke that I got my Our robotics team at the University of
Miami looked a little different than my high
all sorts of whirring, squealing, and start breaking things (robots),
school team. This was the first time I really
thumping coming from the chemistry and then went on to fix things faced the gender gap in STEM and the team
lab. The sounds became impressively (bones). always treated me as an equal, which helped
louder as I opened the door to the As a sophomore in college, me build confidence as an engineer.
classroom. I walked inside and found I applied for my first engineering
all the tables on their side, forming a internship with no relevant work I still work for him many years
sort of gladiator arena in the middle experience on my resume. Instead later, so I’ve had a chance to ask him
of the room. A student welcomed of leaving that section blank, I wrote about that interview. It turns out that
me while gripping a daunting remote about “combat robotics.” I was he had never met a young lady that
controller that took both hands to shocked that every single question built robots. Once I was able to answer
operate. I quickly realized that she was during my interview was related to his questions, he hired me before
controlling a hefty metal machine with robotics. The hiring manager asked I even left the interview room. I’ve
shiny black tires and sharp steel teeth me what voltage I was running, what spent the next 11 years designing and
that was painted like an alligator. battery chemistry I selected, and what developing orthopedic trauma systems
The girls introduced me to Gator motor specifications I chose. I was and I’m still — more often than not —
— their 120 lb combat robot — and thrilled because I probably couldn’t the only woman in the room.
answered all the questions I fired at answer any questions relevant to the Just a couple of years ago, I was
them. By the time I walked out of that position for which I was applying, but I invited to speak at the Girl Powered
room, the open house was over and I could talk about robots all day. panel at VEX Worlds, which is the

SERVO 03/04.2019 43
Contrary to long-held
stereotypes, boys aren’t naturally
better builders than girls. The
difference starts as toddlers, when
they’re handed construction toys
while we’re handed dolls. As a kid,
I would beg my mom to say I was
a boy at the McDonald’s drive-thru
so that I would get a Transformer
robot in my Happy Meal instead of
a My Little Pony plush toy (as I hid
in the back seat so the drive-thru
attendant wouldn’t catch my lie).
It’s no surprise that boys often have
a head start on STEM skills, but it’s
been my experience that — given a
little encouragement — girls have no
BattleBots has given me a platform to share my passion for robotics, and I hope trouble quickly closing this gap.
it inspires young viewers to become builders. Like most kids, I was told I
could be anything I wanted to
largest robotics competition in the because I believed it was possible. I be: a doctor, an artist, an architect,
world. In that room, I saw hundreds of had people in my life that took the even the president! I was already
young girls who not only build robots, time to introduce me to engineering looking at colleges the first time I
but are succeeding at the highest and encouraged me as I developed was told I could be an engineer. I’m
levels of competition. For the first time the most basic skills at my own pace. an engineer in everything I do: my
since I had stepped out of my all-girl I learned to use power tools as a career, my hobbies, my lens of the
high school, I felt I had nothing to teenager, and I was never compared world. I was an engineer before I
prove. to boys that had been using them in knew what engineering meant. As a
Those girls didn’t once question their dad’s garage since they were old little kid, I was always finding ways to
if I was worthy of calling myself an enough to walk. make things with my hands. And yet,
engineer because they were already My mom never told me to stop I almost missed my calling because I
building robots themselves. If they welding when I got home from the didn’t know what to call it. You can’t
could do it, then so could I. It was shop with dirty hands and tiny burn dream to be something you don’t
at that moment that I finally found holes in my high school uniform. Most know exists.
my answer to the question that had importantly, any failures along the way At Season 3 of BattleBots in
stumped me so many times: “How did were an expected part of any learning 2018, only three of the 55 teams in
you decide to become an engineer?” process; they were never attributed to the tournament were led by women.
I decided to become an engineer my gender. While men had 52 shots at the

Girl Powered at VEX Worlds encourages hundreds of girls who are already succeeding at the highest level of
robotics competitions.

44 SERVO 03/04.2019
championship, the world’s Everything is impossible up
opinion of whether women until a “first” proves it’s not.
can build BattleBots was I don’t have to be the
based on the performance of first to fly across the Atlantic
these three teams. or launch into space to have
The gender gap in STEM the power to change our
fields can be a daunting current culture. All I have
hurdle — especially as a to do — all any of us has to
beginner. While I used to be do — is encourage girls when
discouraged that strangers they show interest in STEM
constantly needed convincing activities, and fan that spark
to believe my merit as an until the flame is strong
engineer, I now look forward enough to light its own path.
to these opportunities. It will be a brighter future for
I’m proud to be the first all of us.
female robot builder to meet Soon, the next generation
a skeptical BattleBots fan, will be faced with the same
if answering their questions question I struggled to
means they’ll have higher answer for so long: “How
regard for the next woman Kids watching BattleBots often aspire to start learning did you decide to become
they meet in a STEM field. how to build their own robots. My favorite part of this an engineer?” I hope part
Each woman in STEM photo (besides their excitement) is that she’s wearing of their answer will be that
a princess shirt that reads “Follow your dreams.”
experiences the honor and BattleBots showed them it
Princesses can build robots too!
burden of being a “first.” was possible. SV

www.embeddedARM.com
TS-7180
Single Board Computer

696 MHz i.MX6 UltraLite CPU


4GB eMMc Flash
Up to 1GB RAM
WiFi and Bluetooth Enabled
XBee/NimbeLink sockets
SPI, I2C, CAN, RS-485
Nine-Axis MEMs
Gyroscope, Accelerometer and Compass
Industrial Temp Range (-40 °C to 85 °C)
10+ Year Life Cycle

Board Support Packages,


Source Code and Toolchains
Available

SERVO 03/04.2019 45
MAKE A
SPRING-
LOADED
PEN
HOLDER
FOR YOUR
CNC
ROUTER
46
By Roger Secura

SERVO 03/04.2019
FOR $5!
A good CNC router pen holder can cost up to $75 or more. If you’ve
been thinking of buying one, then maybe you should consider making
one instead. In this article, I’ll show you how to make a spring-loaded
pen holder for five bucks — and thanks to some common plumbing
and hardware parts from Lowe’s and Home Depot, you can make one
in less than a hour.

MARK IT UP
A pen holder is a nice accessory
to add to any CNC hobbyist’s tool
bag. This is especially true for
those just getting started with
CNC routers. For example, rather
than ruining a piece of wood,
the beginner can test a “G-code”
program simply by using a pen
holder to draw the 2D pattern on a
piece of paper.
The pen holder described in this
article (see Figure 1) was designed
with an internal compression spring.
That means the holder’s 3/4” plunge
depth is more forgiving with the
downward movement of the router’s
Z axis.
In other words, if you or a
G-code program make a small
mistake in the height setting of
the Z axis, the pen holder will
have enough travel distance to
compensate for the error.

FIGURE 1 FIGURE 2

QTY DESCRIPTION PART# SOURCE COST


1 1/4” x 1-1/2” Bolt – Hex Drive, AED n/a Home Depot $0.19
1 1/4” Nut-USS, AAB n/a Home Depot $0.07
1 1/4” Split Lock Washer – Zinc, ABE n/a Home Depot $0.17
2 Flat Washer – Brass, SAE #14S Home Depot $0.60
PARTS LIST

1 1” Long Nylon Spacer, I.D. 5/16”, O.D. 1/2” #595 674 Home Depot $0.50
* 1” Long Nylon Spacer, I.D. 5/16”, O.D. 1/2” #65808 Midwest Fastener $0.20
1 1/2” Copper Cap #21728 Lowe’s $0.64
1 1/2” CPVC Male Adapter #50465 Lowe’s $0.89
1 3/8” Copper Coupling with Stop, 7/8” long #505 036 Home Depot $0.98
1 Compression Spring – 5/16” x 1-1/2” x .020” #422418 Lowe’s $0.64
* Compression Spring – 5/16” x 1-1/2” x .020” #88018 Midwest Fastener $0.53
1 BIC Pen – Cristal, non-Clic type n/a Walgreens $0.25
1 Rubber Band 2” long n/a n/a n/a
*Alternative Source

SERVO 03/04.2019 47
To post comments on this article and find any associated
files and/or downloads, go to www.servomagazine.com/
magazine/issue/2019/02.

DEVIL
IN THE
DETAILS
A few notes are in
order before you start
making the pen holder:

• The pen holder’s


length from the tip of
the Bic™ pen to the top
of the 1/4” nut is 4.00”.
That means your router’s
Z axis needs at least
4-1/2” inches of upward
travel distance in order
to install and use the pen
holder.
• The 1/4” bolt threads
on the pen holder may
damage the lining inside
your 1/4” collet (chuck).
FIGURE 3
After using the pen
holder for the first time,
you may want to designate and mark that
collet the pen holder collet.
• The pen holder only works with Bic Cristal
“non-Clic” pens (see Figure 2). Therefore, I
challenge the reader to design a pen holder
that adapts to any type of marker.

CONSTRUCTION
Figure 3 is an exploded view diagram
of the pen holder showing all the prep work
required before the final assembly procedure
begins.
There’s just one critical step that might
require special attention. The hole that’s
required in the 1/2” copper cap must be drilled
in the exact center. In other words, use a
center punch. Also, I found that a small round
file can help to center the hole after drilling.

GET IT TOGETHER
The drawing in Figure 4 and the photo
in Figure 5 show how the pen holder should
look fully assembled.
The only part not mentioned thus far is
the rubber band. Its function is to retract the
FIGURE 4 pen from the plotting surface whenever the
router’s Z axis starts its upward movement.
48 SERVO 03/04.2019
FIGURE 5 FIGURE 6

spring anywhere in my workshop. Maybe the next time I visit


Australia, I’ll look for that stupid spring.
Finally, Figure 6 shows a test drawing from the pen
holder being used for the first time.

BEAUTY AND THE BEAST


The pen holder presented here may not be as sleek of
a design as those $75 versions you find on the Internet, but
It has just enough tension to stop the Bic pen from it’s cheap, easy to build, and it works. SV
falling out of the holder. In other
words, the internal compression spring
pushes the Bic pen downward, and
the rubber band limits how far down THE MOST COMPLETE PROGRAMMING TUTORIAL!
the pen is allowed to go.
• Everything you need to learn to program PIC® micro-
Before you attach the rubber
controllers is included in this package
band to the wood screw and the
tip of the Bic pen, you must twist • Learn the fundamental concepts of programming,
it multiple times to established the including program flow, loops, coding techniques, binary
correct amount of tension. The easiest manipulation, device-to-device communication, user
way to do this is just put one end of interface design, in-circuit debugging, and more!
the rubber band over the small wood
• Includes all the necessary the software, compiler,
screw. Then, twist the other end in $99.99 trainer board, cable, tutorial with exercises and sample
between your index finger and thumb Academic and bulk pricing available
until it’s evenly twisted all along the code!
rubber band. Check out our other popular products
Also, since it only cost $5 to for PIC® developers including:

PBP3
make the pen holder, make sure you U2 Programmer
purchase at least two 1/2” copper PBP3 Compiler
caps. I ruined one copper cap and Prototyping Boards
lost one compression spring when it PICBASIC PRO Compiler Experimenter Boards

www.melabs.com
violently popped out of the holder Contact us at info@melabs.com
during assembly. I never found that
SERVO 03/04.2019 49
Vorpal
Hexapod

By Carol Lynn Hazlett

A ver
hex
STEM
and l
50 SERVO 03/04.2019
What drew me to
look at this robot
was how cute it
is. What made
me buy it is how
versatile it is. Plus,
it has a low price!
To let you know
how much I like
this robot, I bought
five of them! Often,
manufacturers will
send a writer their
product to do an
article on. In this
case, I bought the
robot and liked it
enough to want
to write an article
about it.

ry versatile
T
he name Vorpal comes from
the classic Lewis Carroll
poem, Jaberwocky. The
word Vorpal represents

xapod for a powerful weapon used


to slay an evil monster: the fearsome
Jabberwock. This fits the robot as there
are several combat type games that can

M, gaming,
be played with it.
Vorpal offers several different kits
for sale. There is a basic maker kit which
is just the electronics, and a deluxe
maker kit which has extra sensors, an SD

lots of fun. card reader, and some other items.


These are not bare-bones electronic
kits. All the wiring, connecters, and
hardware needed to assemble the robot
are included. A third kit includes a 3D
printed body. You can choose to print
SERVO 03/04.2019 51
designs for heads
and accessories for it
myself. These are also
posted on Thingiverse.

What is It?
The hexapod is a
six-legged robot with
two DOF (degrees of
freedom) legs. Because
of the unique design,
even though there
3D printed parts for the hexapod and gamepad. are only two servos in
each leg, the robot can
your own or order one from them. make over 60 different kinds of moves and looks remarkably
The files for printing can be downloaded from alive when it walks, dances, or fights.
Thingiverse and are free. I chose the option of printing my There are two elements to the kit: the hexapod itself
own as 3D printing is one of my hobbies. and a controller for it called a gamepad. The hexapod is
Vorpal began as a Kickstarter which occurred in mainly controlled by the gamepad but can be programmed
September 2017 and is open source, so all coding and for autonomous behavior also.
designs are configurable and changeable by the end user The gamepad has an SD card in the deluxe kit which
which is part of the fun of this robot. I have made several allows the user to program a routine using different gaits,
dance moves, and
fight moves directly
from the gamepad,
and then run it.
The brains are
an Arduino Nano in
both the gamepad
and the hexapod.
The hexapod has
a servo controller
module, a BEC
(battery eliminator
circuit), and a
pre-matched set of
Bluetooth modules
for communication
between the
hexapod and
gamepad. A
potentiometer
is used to dial
between modes on
the hexapod.
Power is
provided by
two 18650 li-on
batteries which
are not included.
The kit has the
Hexapod in progress on stand.
holder and proper
52 SERVO 03/04.2019
BEC for using that type of battery.
You’ll need to purchase the right
charger also. Online they do provide
instructions on how to power
the hexapod with other types of
batteries, and explicit instructions on
battery safety. The gamepad uses a
standard 9V battery.
Both Arduinos are pre-
programmed to help you get a fast
start. When updates are made to the
program, they can be downloaded
from the Vorpal site. I use the
Arduino IDE (integrated development
environment) on my hexapods, but
they are also programmed in Scratch
which makes them more compatible
with most STEM programs, and
easier for beginning programmers to
use and learn from.
A trim mode allows the user to
precisely trim the servo positions to One section of Max next to the regular size hexapod.
an exact spot for the more precise
movements from the robot. Most servos can’t be mounted a big enough printer. (They also offer a purchase of just
at an exact angle because of the splines in the servo horn. the big parts if you do have a printer but it’s just not large
The trim mode enables you to overcome that problem. enough for the body and head.)
In the building process, you add some extra jumpers I haven’t built mine yet as the parts for it at the time
which are color-coded and long enough to come out the of this article were in the mail, but I’m certainly looking
accessory port on the front of the
hexapod. This lets you add sensors
of different types to it.
Vorpal now has a gripper
arm available for this port. It’s
downloadable from Thingiverse to
print yourself if you want one.
One of the games for it is a
jousting game. The hexapod has
an add-on “joust stick” protruding
out the front. Mounted on the stick
is a small object attached with a
small magnet. Each hexapod tries
to knock the other hexapod’s object
off the stick. It sounds simple, but
it’s a competitive game which takes
a bit of thinking skills.

Bigger is Better
Max is here! The same cute
little hexapod but twice the size!
I’m in the process of 3D printing the
parts which you can also buy already
Servos installed; getting ready for the electronics.
made from Vorpal if you don’t have
SERVO 03/04.2019 53
Building session in progress.

forward to it. As much excitement as the little hexapod which is easy enough to put together for beginners, but
causes when I demonstrate it to other people, “Max” is sophisticated enough to keep more experienced users
going to be a real show stopper. There are some videos of interested and challenged.
this guy on the Vorpal website at vorpalrobotics.com and There are many ways to use this little hexapod and
on YouTube. many changes or additions a user can make. With the open
source programming and 3D printed parts, there are endless
In Conclusion possibilities for creating new versions or games for it.
Like all things in life, it’s not perfect but in the five of
them I bought, there were no bad parts or failures. The
Vorpal has come up with a fun and cute hexapod customer service is excellent, and all the complaints or
errors are shared with the entire
community in the forum or by email.
I found every question I had
answered in the building and
assembly instructions. Vorpal has put
extensive guides and manuals on
their website, and is expanding and
updating them all the time.
If a subscriber to the forum
wants to be on the email list, all the
members of that list can read the
questions and answers from them.
I like it!
Several questions I had were
answered by reading these answers
to other people’s questions.
As a new startup, Vorpal is
going in the right direction. They
have some wonderful products that
stand out from the crowd of robotic
startups that are popping up all
over.
As you may have guessed, I’m
Gamepad in construction.
enjoying my Vorpal purchases! If you
54 SERVO 03/04.2019
Holiday versions.

To post comments on this article and find any associated


have any questions, feel free to email me at carolhaz391@ files and/or downloads, go to www.servomagazine.com/
gmail.com. SV magazine/issue/2019/02.

Introducing the Extreme Robot

1(:39-U
Motor Control
www.robotpower.com
Scorpion Ultra
The BIGGEST 6 7V - 36V
2x - 70A Peak H-Bridges
Scorpion EVER
0LQL&LUFXLW%RDUG+ROGHU
6
Big Power! 6 50A continuous w/cooling
6 Current/Temp protected
Big Bots! 6 R/C inputs w/mixing
6 4.50" x 2.35" x .7”
6 Enclosure available
'(6,*1('
)25 29(5
Big Motors!
7+(('*(
&20321(17
%2$5'6 MADE IN
Model The Tiniest
220 THE USA Scorpion EVER
Scorpion Nano
6 4V - 16V
Innovative H
Holding
olldi
o din
ngg SSolutions
olutio
olutiio
ut ons
n 6 2x - 4A Peak H-Bridges
6 2A continuous
Current/Temp protected
800-759-7535 6
6 R/C inputs w/mixing
6 18.5mm x 15mm x 5mm
6 ~5g
Phone: 360-515-0691 s sales@robotpower.com

SERVO 03/04.2019 55
Balancing
a Robot
on a Ball
By John Blankenship

Assume you have a mobile robot sitting on a large ball.


Imagine programming the robot so that if the ball is
moved, the robot will climb to the top of the ball and level
itself. If building such a robot excites you, this article will
demonstrate some key points to help get you started.

W hen it comes to balancing robots, a lot has


been written on two-wheel Segway-like robots.
They can be both interesting and challenging
for sure, but if you have read some of my
previous articles, you know I often like to find
a different approach to standard topics.
My inspiration for a different kind of balancing project
came from the Star Wars BB-8 shown in Figure 1. When I
first saw pictures of the robot, I assumed the large ball was
head is not the controlling mechanism.
A search of the Internet can reveal details of the inner
workings of the BB-8. The wheeled propulsion mechanism
sits inside the body at the bottom of the ball. This assembly
can move the ball in any direction, but it can also rotate
and tilt a curved arm that rides near the inner surface of
the upper half of the ball. The head attaches to this arm
magnetically allowing the head to move totally independent
of the body’s orientation.
being moved by wheels on the head. If you watch the robot Building a BB-8 using an internal mechanism should be
in action, though, you see that the head can tilt far to one within the talents of many hobbyists because the low center
side when the ball is sitting still (as if the robot is peeking of gravity of the propulsion unit should make programming
around a corner, for example). Obviously, this means the relatively simple. For that reason, I was more intrigued with
56 SERVO 03/04.2019
To post comments on this article and find any associated
files and/or downloads, go to www.servomagazine.com/
magazine/issue/2019/02.

Figure 1. Figure 2.

the possibility of placing a typical mobile robot on top of a


ball and letting the movements of the mobile robot control
the movement of the ball.
Figure 3.
The first step in achieving such an action is to prove that
a mobile robot can constantly level itself on top of a moving
ball. Figure 2 shows the robot in action.

Constructing the Hardware


The robot shown in Figure 2 is one I had used for a
previous project. It’s powered with 360-degree servo motors
(Figure 3) and controlled with an EZ-Robot controller
(Figure 4). The controller had numerous ports, an integrated
battery, and Internet communications, so it made it easy to
implement a simple feasibility test.

Sensing Tilt
In order to balance atop the ball, the robot must be
able to determine its tilt orientation. I used an ADXL 335
accelerometer with analog outputs as shown in Figure 5
to obtain the robot’s orientation. The ADXL 335 provides
an output voltage for each axis that is proportional to the
actual tilt. If you look carefully at Figure 4, you can see this
sensor attached to the front of the EZ-Robot controller.
The sliding caster assemblies originally used to maintain
balance on the mobile robot I used had too much friction
SERVO 03/04.2019 57
Figure 4.

against the ball’s surface, so I replaced them with wine corks Robot has the option of programming their controller with
(one on the front and another on the back as shown in a primitive version of Blockly. This made testing extremely
Figure 6). Each cork is tipped with one prong from a plastic easy, but I will probably switch to a Parallax Propeller (and
fork (hot glue is wonderful). The tiny points at the end of Parallax’s more advanced Blockly language) if I decide to
the prongs move easily over the surface of the ball. explore a more sophisticated version of this project.
I will confess that I wasn’t sure how difficult it would be
Programming the Robot to create a self-leveling program. Figure 7, however, shows
a very simple program that works far better than I expected.
After the hardware construction was completed, I was The robot should be placed in a level position at the
anxious to find out how hard it would be to program the top of the ball before starting the program. This is required
robot to maintain its position on the top of the ball. EZ- because the program starts by reading the tilt values
from the accelerometer and stores them in the variables
LevelFront and LevelSide to be used as level reference values.
Figure 5. Next, two variables are initialized to the center (stopped)
values for the wheel servos. Subsequent tilt readings from
the accelerometer will be used to control the wheel speeds
by deviating the servo pulses from these specified center
values.
The endless loop in Figure 7 obtains the current
forward and side tilt values from the accelerometer and
stores them in CurFront and CurSide. In each case, the new
value is compared to the stored reference values for a level
robot, and the difference is stored in the variables FrontDiff
and SideDiff.
Notice that each difference value is divided by a scale
factor (obtained from experimentation) to create a suitable
value for the next operation.
The FrontDiff is added to the center value for one
wheel’s servo and subtracted from the other. The reason
the value is added to one and subtracted from the other is
because the wheel servos are mirror images of each other
58 SERVO 03/04.2019
(meaning to get both motors to move the robot
in the same direction, the pulse to one must
be increased while decreasing the pulse to the Figure 6.
other).
The net result of this action is that both
motors will either move forward (or backward)
at a speed based on the value of FrontDiff. This
just means that the further the robot is tilted
forward (or backward) on the ball, the faster it
will move, trying to get back on top.
The value of SideDiff is used in a similar
way except that it’s added to both wheels (or
effectively subtracted if the difference value is
negative). This means that one wheel will move
forward and the other will move backward,
causing the robot to rotate around its center
until the side tilt is eliminated, effectively
ensuring that the robot is facing the top of the
ball. At that time, the robot’s forward tilt will
cause it to climb to the top of the ball.
It’s important to emphasize that this
algorithm changes the wheel speed based
on the difference in the robot’s current tilt
compared to the desired level position — the
more tilt, the faster the robot moves. This means
the robot’s speed will be proportional to the
error between the current tilt and the desired
level position.
Since we’re only using the proportional
aspect instead of full PID (proportional, integral,
differential) control, the program is not good
enough to make the robot balance on a fast-
moving ball. I found it very pleasing, though, to
see the robot easily level itself at the top of the
ball as I manually rolled it around at moderate
speeds.

Final Thoughts
The program in Figure 7 demonstrates
that a mobile robot can level itself on a slowly
moving ball. The program has its limitations,
but it certainly demonstrates the feasibility of a
ball-balancing robot. The next logical step is to
implement a more rigorous algorithm using a
more capable language.
Once the robot can stay balanced on
a freely moving ball, it should take minimal
effort to make the robot move the ball around
under remote control. Interested readers are
encouraged to explore this challenge. SV

Figure 7.
SERVO 03/04.2019 59
Giving Double
Jeopardy an
Unconstitutional
Amount of Shots
By Bryce and Evan Woolley

Back from BattleBots 2018.

60 SERVO 03/04.2019
Twin brothers hack whatever’s put in front of them,
then tell you about it. Twin Tweaks
By the time you’re reading this in the pages of SERVO
Magazine, it will be public knowledge that Double Jeopardy is
returning for the 2019 season of BattleBots™. It will still be the
best (and only) cannon in BattleBots’ history, but our ultimate
goal with the robot is not simply novelty. We truly believe
that cannons can become competitive in combat robotics, and
this year we are taking huge (maybe even mammoth) strides
towards that goal.
Instead of one
major upgrade
this year, we are
angling for two: a
multi-shot cannon
and a more robust
design overall. Are
we chomping off
more than we can
chew, destined for
disappointment
and a tantrum?
Or, will DJ’s new
upgrades have our
opponents begging
for a witch doctor?
There was only one
way to find out.
SERVO 03/04.2019 61
To post comments on this article and find any associated files and/or downloads,
go to www.servomagazine.com/magazine/issue/2019/02.

Even though we consider the 2018 season a


smashing success, we knew there were plenty of
upgrades that would help Double Jeopardy stand a
better chance against the mechanized rampage of
the opposing BattleBots.
The most obvious area of improvement was
the cannon. In our 2018 design, the goal was to
make the strongest cannon possible to prove that
a cannon could actually do damage to competing
bots. Our 2018 cannon — made from schedule
80 steel pipe and schedule 160 threaded steel
fittings — had a maximum operating pressure of
3,500 PSI. For various reasons, we ran the cannon
at 1,000 PSI that season, but we still did damage
even at that reduced pressure. So, we have more
than enough pressure and nitrogen capacity in the
cannon, and the best way to improve its damage
output would be to increase the number of shots.
Easier said than done, of course.
Our other goal — which we considered just
as important as achieving multiple shots — was
to make the bot more robust. Our frame in 2018
was made from bolted together 6061 aluminum
— not even remotely sufficient for the rigors of the
BattleBox. In one sense, it had worked as we had
hoped; the thin panels and soft frame were ripped
The spring steel torsion spring mechanism. apart by the horizontal spinners of our opponents,
immobilizing Double Jeopardy without destroying
Post-Trial Briefing the expensive insides. There was no need for the Four
Overall, we were thrilled with Double Jeopardy’s Horsemen to keep horsing around once they disabled half
performance in 2018. The cannon worked both times; we of our drive train. However, we want to do more than make
had 100% accuracy; and we did damage in every match. For a glass cannon that can take one shot then get ripped up.
all the gory details, you can catch up on every episode of the A cannon that could take two shots and then get ripped up
2018 season on Amazon Prime Video or the Discovery Go wouldn’t be much of an improvement.
app. Fortunately, these goals were complimentary. One of
the improvements we had in mind
for the cannon was to use a welded
tank. With a welded tank, we could
use schedule 80 butt weld fittings
which are significantly lighter than
the schedule 160 threaded versions.
The weight savings would account
for both the slight extra weight for
the multi-shot actuation mechanism,
as well as free up some weight for a
real welded steel frame with thicker
titanium armor.

Requests for
Production of a
Better Tank
Our 2” Dynaquip valve. Before we get down to the
62 SERVO 03/04.2019
details of the build, we want to make something
crystal clear: Building robots can be dangerous.
Building combat robots can be especially dangerous.
Building a 250 lb robot cannon can be super-duper
dangerous. Articles like these — where we describe
our design and build process — are meant to be
inspiring in the general sense and hopefully helpful
in the specific sense, but these are not meant to be
step-by-step guides.
We take ample safety precautions at every
stage of the build and testing. Descriptions of every
step of our safety procedures or vivid and tactile
descriptions of the PPE we use at every step would
be too voluminous to include in every article. So,
please take what design inspiration you can from
our odyssey to make the best cannon in BattleBots,
but if you set out to build something, please take
proper safety precautions.
Sophisticated welding is not unusual to see in
BattleBots, but making a custom welded pressure
vessel presents a unique set of challenges. None of
our team members are passable welders, but even Plotting our upgrades.
if we were, the cannon would still be well above
our pay grade. Only an ASME certified welder should craft heat treater, the tank looked a bit like a scaly mess, but that
a custom high pressure vessel, and certainly nobody on the rough exterior gave us the confidence that the unit would
team was ASME certified. pass the next step of the process: the hydrostatic test.
Fortunately, we found Dean’s Certified Welding Pressure vessels are often put through a process called
in Double Jeopardy’s
hometown of Temecula,
CA, and they did a fantastic
job with the tank at a great
price. They even provided
valuable feedback on the
design, including the advice
that instead of a tee fitting
between the legs of the
tank and the valve, we
should use a section of
pipe cut to integrate with a
special fitting. This actually
helped us avoid using a
coupler between the tee
and valve, saving weight
and allowing more of the
center section length to go
towards the barrel.
After welding, the tank
needed heat treating to
relieve stresses in the unit.
Without heat treating, the
tank could split apart at the
welds and we would have
to start from scratch all over
again. Once back from the Gutting DJ 2018.
SERVO 03/04.2019 63
we had the tank pressured to 6,000 PSI, and it held
fast like the Kraken chomping down on Red Devil. The
tank passed the test, and we officially had a pressure
vessel rated for an operating pressure of 3,000 PSI.
Even after all that, we were nowhere near done
with preparing the tank for the robot. After the
welding and heat treating, both the tank and the
welded multi-shot barrel were still scaly messes. The
barrel in particular needed some attention to ensure
that the inner surface was smooth and provided no
obstructions to the travel of the slug. The welds on
the tee fitting serving as our reloader fully penetrated,
which is great for the integrity of the welds and
the strength of the unit. However, that also meant
the inside surface of the reloader presented a bit of
terrain that impeded the slugs.
We started with the bodged-together flapper
stick that we used for the basic cleanup on our
barrels, but progress on that was as slow as
BattleBots’ journey from Comedy Central to ABC. We
Laying out the new drive train. couldn’t spend the entirety of our precious build time
smoothing out the barrel, so we acquired a ball hone
hydrostatic testing. Hydrostatic testing is a method used to make quicker work of the task — more at the pace of
for ensuring that pressure vessels can handle their working BattleBots’ journey from ABC to Discovery.
pressures, and is often used for units like scuba tanks. The
process basically involves submerging the unit in water,
then filling it up with water at pressure to see if the unit will
Going with the High Flow
burst. The test pressurizes the vessel to a predetermined Last time, we described the long design road that
multiplier of the operating pressure based on the factor of eventually led us to a Kinetrol rotary vane actuator as the
safety. solution for the new cannon valve actuation mechanism.
If the unit successfully holds pressure, it passes. The Double Jeopardy uses a 2” Dynaquip ball valve in the
maximum onboard pressure allowable under the BattleBots cannon. The valve is rated for about 6,200 PSI and has a
rules is 3,000 PSI, and we wanted to ensure that we were very high actuation torque at higher pressures like 3,000 PSI.
capable of running at that pressure. For the hydrostatic test, The actuation torque is around 100 ft-lbs, and generating
that kind of torque in a
split second is a major
challenge.
The Kinetrol rotary
vane actuator is an
effective and elegant
solution, able to
generate 1,740 in-lbs
(145 ft-lbs) at only 100
PSI.
Even better than
the efficacy of the
mechanical solution,
however, was the
help from the team at
Kinetrol USA. They are
BattleBots fans, and
were true partners in
the design and build of
Testing the multi-shot barrel prototype.
Double Jeopardy 2019.
64 SERVO 03/04.2019
The Kinetrol team gave us pointers on how to
achieve the actuation speed we wanted, provided
feedback on the best way to mount the actuator
to the valve, and even modified the stock unit for
us. The standard actuator comes with 1/4” ports,
but Kinetrol USA took them out to 3/8” ports. This
increased the flow and thus the actuation speed.
Even with all the help from Kinetrol, there were
still some mysteries for us to solve. We wanted the
fastest actuation speed possible. A fast actuation
speed would improve the acceleration and ultimate
muzzle speed of the slug because it would allow the
maximum pressure to be accelerating the slug for
the longest time.
The spring steel torsion spring from last season
worked to open the valve, but it wasn’t the fastest
and it didn’t have a nice even opening speed
throughout the arc.
We think the valve opened slower than it could;
so, as the ball valve turned, it only let out a fraction
of the pressure at first. This started pushing the slug
down the barrel at low pressure and acceleration,
and once the valve opened completely there was Wood-aided design mock-up of the new frame.
only a little more length in the barrel over which the
slug could accelerate at max pressure. the highest flow system possible, and at high Cv values we
Fast actuation speed was also essential for our double would achieve actuation speeds that were a small fraction of
shot mechanism. We needed to open and close the valve so a second.
quickly that it closed before the
second slug fell down the gravity
magazine. If the actuation was
too slow, the second slug would
go flying out right behind the
first. So, getting that actuation
speed right was essential.
The key consideration
for maximizing actuation
speed was to maximize flow;
operationalized using the
flow coefficient Cv. The flow
coefficient describes the
efficiency of flow through a
system. The Cv itself is a number,
and it means the volume (in US
gallons) of water at 60°F that
will flow per minute through a
valve with a pressure drop of
1 PSI across the valve. So, for
our purposes, a higher Cv was
better.
Kinetrol helpfully provided
stats on the actuator showing
what the actuation speed would
be for various flow coefficients.
It was clear that we needed Beautiful welds on a new barrel.
SERVO 03/04.2019 65
the system had high flow coefficients.
The pneumatic subsystem for the
Kinterol was pretty basic: We had
an air tank; a solenoid valve; and
the Kinetrol actuator. The air tank
supplied the air for the Kinetrol. The
actuator required about 32 cubic
inches of air per stroke. Since we
needed to open and close for the first
shot, then open for the second shot,
that meant about 100 cubic inches
was all we needed.
To keep things as simple and
as high flow as possible, we used a
100 cubic inch paintball tank that we
pressured to 150 PSI, which was the
maximum pressure for the actuator.
The regular operating pressure for
the unit is 100 PSI, and our sponsors
warned us that upping the ante to
150 PSI would shorten the cycle.
However, we didn’t need
anywhere near the millions of cycles
we would get out of a 100 PSI
system. We had a need for speed, so
The new welded tank. Ready for hydrotesting. 150 PSI it was.
By running the whole subsystem
To ensure the Kinetrol subsystem had a high flow on 150 PSI, it precluded the need for a regulator. If we had
coefficient, we needed to ensure that all the components of a regulator, we could potentially use a smaller volume tank
at higher pressure; regulate the output
of the tank down to 150 PSI; and go
from there.
We ditched the regulator for two
reasons: We couldn’t find a good
regulator to get us from high pressure
to 150 PSI; and because all the
regulators we looked at were very low
flow. It would hamstring our efforts to
make a high flow system, so a big tank
and no regulator was the way to go.
The other component of the
system was a little trickier. To operate
the Kinetrol actuator, we needed a
solenoid valve; specifically, a four-way
solenoid valve. Four-way solenoid valves
are used when you have a pneumatic
system that you want powered in both
directions. In our case, we wanted a
pressurized open and a pressurized
close.
Finding a good high flow, four-way
solenoid that didn’t break the bank or
our weight budget was a challenge.
Cleaning up the welded barrel. An option that looked promising
66 SERVO 03/04.2019
was a fairly high flow metal
body solenoid valve. It was a
little heavy, but the real last
straw was the price. It was a
specialized piece normally used
in nuclear reactors and cost
over $600.
There had to be a better
way ... and there was: We
found a great high flow four-
way solenoid from Parker: the
B612BB549A. A 3/8” port, 24
VDC, four-way solenoid valve
with a commendable 2.7 Cv
that only cost about $150. A
higher Cv would have been
even better, but this was not
bad at all for a 3/8” port.
To ensure that we knew
how the Kinetrol subsystem
would work, we built up a
test bed. We included gauges A Kinetrol rotary vane actuator.
coming off of every line, so we
could read the pressure coming from our tank, through the automatically when we released the switch back to the zero
solenoid valve, in both lines going from the solenoid valve to position. That meant we didn’t have to think too hard about
the Kinetrol actuator, and even out the exhaust ports of the timing both the open and close. It was all just one flick of a
solenoid valve. switch — the perfect way to put our opponents in jeopardy.
For ease of
experimentation, we gave
our test bed its own battery
and hooked up the solenoid
valve to a simple DPDT
switch that we thought
provided a nice analog to
the throttle stick on our
transmitters. We pressured
up the system with our
trusty nitrogen cylinder,
monitored the pressure on
all our gauges, and let it rip.
The Kinetrol actuator is
satisfyingly loud, making a
nice wooshing sound and
a clack as the vane thrusts
back and forth. We were
also thrilled with how the
solenoid valve operated on
the switch.
Flipping the switch
opened the actuator, and it
held open for as long as we
held the switch. What we
particularly loved was that
the actuator would close Hooking up our pneumatic test bed.
SERVO 03/04.2019 67
more robust 6” Colson wheels
instead, and sacrifice invertibility.
This also gave us some flexibility
with the shape of the robot.
By only having 6” wheels to
enclose, we were able to make
the outer portion of the frame
shorter, and angle it up towards
the center of the robot with
the cannon perched on top.
We thought it made for a much
cooler look.
We started by mocking
up the frame in wood, and
then longtime friend of Robot
Central, Darrell welded up the
frame for us. The frame clocked
in at a little over 50 lbs; about
30 lbs heavier than the 2018
joke of a frame. That was still
less than the weight we saved
out of the cannon, so it looked
like we were on track weight
wise.
The new cannon was
welded and the actuator
Proof of concept! working flawlessly on our test
bed. However, even with a
Who Framed Double Jeopardy? robust new frame welded up, there was still a lot to do.
We had to design and cut the floor; assemble the
Simultaneously with our work on the new cannon, we entirely new drive train around the 6” Colson wheels;
designed a new frame for DJ. The majority of the weight transmogrify the Kinetrol system from a test bed to a battle
savings from the welded tank were invested in a welded hardened in robot system; sort out the fill and purge systems
steel frame that would allow the bot to take a hit. The on the tank; put it all together; wire it up; make sure it
whole process of having the tank welded, heat treated, and worked on the radio; and see if we could do it all within our
hydrotested took a couple of months, so we took that time 250 lb weight limit.
to plot out the best way to make a strong frame even out of Don’t worry! By the time where we got in the build to
fairly thin walled tube. where we are in this article, we had all of about two weeks
One major design driver was the new valve actuation before the competition. Would that be plenty of time, or
mechanism. Last season, our spring steel actuator was were we actually the ones in jeopardy?
situated underneath the cannon, allowing the cannon to sit Find out next time in SERVO Magazine, and on the
on top of the robot and give the whole design a distinctive Discovery Channel starting June 7th! SV
look. We always operated on the assumption that we
would do the same thing this year: Situate the actuation
mechanism under the cannon and go from the there.
The Kinetrol actuator, however, was taller than the
spring steel arrangement by a few inches. This raised the
overall height of the robot from the floor to the top of the Special Thanks
cannon to about 11”, and that wasn’t even taking into
account the height of our reloading mechanism. Phil McKinney from Kinetrol USA
This meant that even the large 10” wheels we used
last season would not be tall enough to make the robot Darrell Hellofs from Virgin Orbit
invertible. That, however, ended up being an easy design
decision. We opted to ditch the 10” wheels, go with far
68 SERVO 03/04.2019
To Order Visit www.servomagazine.com
or call 1-800-783-4624
Some Of Our Most Popular Books On Robotics
Arduino Robot Robot Building Robot Programmer’s The Ultimate Guide
Bonanza for Dummies Bonanza To DIY Animatronics

Nuts & Volts and SERVO On CD

SERVO 03/04.2019 69
Service Robots
are Knocking on
Consumer’s Doors In my 175+ articles in this
Obviously, I’m and other magazines, service,
not talking about home, and personal robots
a bipedal robot comprise most of what I write
about. For those who watch
standing at a door and read science fiction movies
with a curled- and stories — from Forbidden
up fist, literally Planet, The Jetsons, Asimov’s
pounding away iRobot series of stories, Johnny
5 in Short Circuit (shown in
on the door. Figure 2), and even cute little
However, that’s R2D2 who squeaked and
not really far off beeped his way into our hearts
from what many in Star Wars — a robot in our
homes sounds just right.
people think home
Service robots are
robots should be frequently divided into three
like. categories, with the category
Figure 1 - KAIST Hubo kneeling at the door during the DARPA of ‘service robot’ usually
Figure 1 shows the winning competition. ascribed to automatons that
robot, Hubo (designed by serve multiple people, not just
Korea’s Team KAIST) that one person within a home.
participated in DARPA’s The ‘home robot’ is usually
2015 Robotics Challenge. It’s the title for any bot that
not really knocking at the works within a home and can
door while on its knees; it’s perform tasks for one or many
completing one of the many people, or just be programmed
tasks incorporated into the to do routine tasks within
competition. Hubo won the a home like daily vacuum-
first prize of $2 million at the cleaning. The ‘personal
event for completing several robot’ is usually a robot that
simulated disaster-assisting is designed to assist a single
tasks within a few minutes, person in daily living. However,
but it took many years of all three terms are frequently
research and development used interchangeably by many
to build the robot. This very people.
complex and costly robot
may not be the home service
robot for today, but its unique
Successful and
technology could easily be Failed Robot
simplified to provide a viable
service robot for tomorrow’s Businesses
home use. Figure 2 - Ally Sheedy dances with Johnny 5 in ‘Short Circuit.’ Quite often, movies and
70 SERVO 03/04.2019
Then NOW

and
Advances in robots and robotics over the years.

By Tom Carroll To post comments on this article and find any associated
files and/or downloads, go to www.servomagazine.com/
TWCarroll@aol.com magazine/issue/2019/02.

Figure 3 - iRobot PackBot. Figure 4 - CyPhy Works


tethered drone.

books with robots used around people media stating that “this robot is truly a similar degrees in our nation’s many
sets one’s mind reeling — especially game changer.” top technical universities and have
for those of us who have a technical The annual Consumer Electronics left to form companies of their own,
bent to actually go out and design Show (CES) in Las Vegas this past but then, reality sets in: Just how do I
and build a robot ourselves, and January was the place for any large design, build, and sell a robot for use
then attempt to market it to others. or small electronics manufacturers with humans in their homes?
However, it takes more than a good to be that wanted to present their Most utilize innate talents
idea to make a product successful in exciting new products to the media that can’t be taught, and go on to
the marketplace. and potential buyers. In the last formulate truly miraculous products
It has often been said that decade or so, personal and home through much perspiration. iRobot’s
inspiration is only about 5% of the robot manufacturers have been taking Roomba robot vacuum cleaner jumped
effort and perspiration 95% of what a larger and larger presence at this ahead of all the others who came later
it takes to bring forth a product world-class exposition, with many and has sold tens of millions of these
to the consumer. Thoughts can be different types of robotic products units.
exceptionally great, but the most taking center stage. Getting out there first with an
effort goes into the transformation of The annual attendance is entirely new product can be key.
those thoughts into a viable robot that particularly critical for newly-formed However, just being first with a new
offers capabilities people will actually ‘robot’ companies that have struggled type of product doesn’t always work.
want to purchase at a reasonable for several years, begging for funds After the formation of iRobot
price. through Kickstarter crowd-funding and and the development of the wildly-
That’s certainly the case with other venture capitalists. Usually, small successful Roomba vacuum cleaner,
designing, building, and selling robots companies can’t afford the splashy TV Greiner’s greatest talents showed
to the general public. One can have and other media ads, and count on through with the development of
the greatest idea of how a certain the media present at CES to hopefully military robots such as the PackBot
robot design will “sell like hotcakes,” praise their new robot. shown in Figure 3. Greiner left
only to find out that several years of iRobot in 2008 to form the tethered
many trusting investor’s funds have drone company, CyPhy Works which
‘gone down the drain’ and the only The Success of iRobot later developed a platform called
result is a series of slightly improved Helen Greiner, one of the co- Persistent Aerial Reconnaissance and
robot prototypes that still can’t meet founders of iRobot, often tells Communications (PARC), among other
basic functionality and price point audiences: “It was my extreme interest projects. Their tethered drone concept
targets. in Star Wars’ R2D2 that led me into is shown in Figure 4.
Most of the recent failures looked MIT and receiving a bachelor’s degree Greiner has since left CyPhy Works
so spectacular in the beginning in Mechanical Engineering and a and is now working with the US Army
that they took center stage at many master’s in computer science.” Many on robotics, autonomous systems, and
expositions around the world, with the other talented graduates have earned artificial intelligence initiatives.
SERVO 03/04.2019 71
that their recent purchase by the
huge German company, Bosch was a
blessing as it appeared that continued
funding would assure that the $800
Kuri (shown in Figure 5) would easily
go through its development and
manufacturing stages and soon find
its way into homes around the world.
However, this elation was short-lived
when the financial people at Bosch
realized that Mayfield’s glowing
financial sales figures were well off the
mark.
Cynthia Breazeal — who developed
the social robot called Jibo — was (and
is) a world renown robotics engineer
and scientist who certainly knows the
technical aspects of robotics. You can
see the complexity of the Jibo robot
in Figure 6 that might give a hint of
the difficulties in the development of
the robot from concept to sales to the
masses.
Her true goal for her robot was to
be a social companion; in particular,
for the growing demands of senior
Figure 5 - Kuri home robot in the kitchen. citizens. Despite the complexity of the
multiple axes of motion in the base,
Recent Failures category of home robots. Quite this robot could just wiggle itself into
in Home Robot frankly, the little robot could only
roam around and do very little else
the hearts of those nearby. Breazeal
counted on the intelligent voice of Jibo
Businesses in a manipulative way. The images of to provide social companionship to a
CyPhy Works rebranded as Aria arms (or as I see them, flippers of a user.
Insights in January of this year and penguin) are just that — images and Her many years of campaigning
then disappeared in a puff of smoke not moving appendages. for funds to finance the design,
two months later. Their robot could Mayfield’s management felt construction, and marketing of Jibo
have ‘knocked on your door,’ allowed larger corporations to
but like many others, is no bring forth to the marketplace
longer in existence. Too a unique series of similar
many people have found Internet-connected devices at
this sector of robotics to be far lower costs than her $899
a fickle and volatile business price tag. Jibo didn’t have any
venture. Unfortunately, more useful movements other than
companies will go out of swiveling its head to look at a
business this year. human talking to it.
In previous articles, Amazon’s Alexa-capable
I discussed the cute little devices and Google’s Home
Kuri developed by Mayfield digital verbal assistants (in my
Robotics that was a huge opinion) seemed to stomp on
hit at last year’s CES. The Jibo’s rise to fame with their
diminutive mobile robot was far lower cost. These billion-
set to take the home robot dollar corporations could
world by storm as a social funnel millions of dollars into
robot — a recently-developed Figure 6 - Jibo talking with his “undressed” brother. development in a far shorter
72 SERVO 03/04.2019
time than Breazeal’s small company their homes when they could use a many Goliaths out there.”
could. She ultimately spent $70 million smart device like a Google Alexa to
but it was to no avail. retrieve any sort of information they
Though Jibo was once heralded might desire, is beyond me. Stumbling Blocks
as “the first social robot for the Unfortunately, so many seemingly in Home Robot
home” and was named one of Time bright ideas for a home robot run
Magazine’s “Best Inventions of 2017,” the same paths as Keecker when the Development
Breazeal finally had to call it quits developer and buyers suddenly come Before discussing the types of
last year. Jibo recently turned off its to the same conclusion: “Why are we robots that might be appropriate for
servers, which restricted existing users trying to stuff basic entertainment a home environment, I’d like to look
from accessing software updates. This and information systems into a robot at some of the potential obstacles
will severely limit the functionality of platform?” that robot manufacturers might
the robots in the field. Fortunately, Pierre Lebeau, founder and CEO encounter in their marketing plans.
Mayfield Robotics shut down Kuri of Keecker, made the announcement The US government and the numerous
before it was shipped to consumer’s of ceasing business in his blog. Lebeau rules and regulations seem to be the
homes. said Keecker “ran out of cash” as it biggest hurdle for potential sellers —
Another promising social robot for was unable to find investors to keep especially for robots that might assist
the home user was Keecker, produced the company going. “Building a elderly and physically disabled persons.
by a consumer robotics company profitable cash machine in consumer A robot running around the floor of
based in France. Unfortunately, it has robotics is harder than we thought, a typical home (such as a Neato or
joined the ranks of Jibo and Kuri, and and we couldn’t convince investors to Roomba robot vacuum cleaner) could
also ‘bit the dust’ this year. further fund us.” cause someone to trip, but so could a
Keecker was formed in 2012 Lebeau commented the company child’s toy on the floor. The FDA and
and managed to raise $8 million — had “over 1,000 robots out there, other government agencies are more
enough to produce and distribute thousands of users, and have seen concerned with personal assistant
1,000 robots to customers. The some amazing stats like 3.5 hours robots that are intended to physically
‘HomePod’ (shown in Figure 7) could of usage [per] day.” Keecker didn’t assist or handle a person with limited
autonomously navigate and provide receive as much attention as Jibo and strength and movement capabilities.
users with entertainment, A prospective seller
security, communication, of a home robot might
and smart home data. sit down at the beginning
Keecker ads spoke of the with just a glowing idea.
uniqueness of their bot “This robot will be a hit
this way: and I’ll sell millions!” He
“From entertainment or she might look at costs
and games to web and the resulting funding
browsing, from media issues, and start to see
immersion and digital that the road ahead will
decoration, to home have several hurdles.
security, Keecker is the Bringing in the talent
world’s first HomePod required for the different
made to inspire your aspects of a new robot
interior. It’s a totally company makes the
new type of ‘computer’ Figure 7 - Keecker in the living room. “dreamer” realize that he
entirely designed with or she must have some
the home in mind to share real-life Kuri, but it suffered the same fate, human resources knowledge.
moments, share content, and share sadly. Lebeau further stated “We’ve When the new company team is
experiences.” seen others like Jibo and Kuri fail, we starting to grow, the dreamer must
In my humble opinion, fought like dogs not to follow their take on top management tasks or hire
Keecker seemed to be just a home path, but in the end, there is a strong a CEO for the job.
entertainment system with a video sentiment against consumer robots However, I think the greatest
projector crammed into a mobile that we couldn’t get past. In the end, obstacle really is (as I mentioned
robot’s body. Why people would want consumer electronics is a tough market earlier) the fickle potential buyer of
something fairly large roaming around where David is likely to fall against the the home robot — a very difficult
SERVO 03/04.2019 73
times more powerful and thousands
of dollars cheaper than his room-sized
clunkers.
What happened with producing
viable home computers for consumers
is really opposite from what is
actually happening in the home and
commercial robotics businesses. At
the dawn of the personal computer
revolution, people like Steve Jobs and
Steve Wozniak or Bill Gates and Paul
Allen were just salivating at the chance
to build an affordable computer for
the masses and supply an endless
series of software applications. They
certainly had heard about robots, but
they jumped at the immediate needs
that they observed in small businesses
and homes. In those days, only SciFy
enthusiasts and dreamers could
Figure 8 - DEC VAX 11-780 computer. envision a robot in the home.

person to understand indeed! vision. In my previous article on robots


for the masses, I mentioned a quote What Makes a Robot
The Reality of Selling from Ken Olsen in 1977. Olsen was Different from a
the founder of Digital Equipment
a Home Computer Corporation (DEC) that made mid- Computer?
sized business computers such as the People like to place the home
Versus a Home Robot PDP-11 and VAX minicomputers, of computer and the home robot in the
In my last column, I wrote about which the very popular VAX 11-780 same category as ‘high tech’ products
“Robots for the Masses” and discussed is shown in Figure 8. He stated: to be used by non-technical people in
what technologies had developed over “There’s no reason anyone would their everyday lives. However, they are
the years to allow home robots to be want a computer in their home.” I’m not at all alike. A home computer is
truly functional and safe around us not sure why any executive whose a non-moving information processing
fragile human beings. A large industrial company builds a certain product for and retrieving device, whereas a
robot might be able to rapidly handle larger businesses would not think ‘out home robot is a moving, sensor-laden
and manipulate a 20 pound object, but of the box’ for a moment and perceive electro-mechanical device with built-in
that speed could also potentially kill of new market areas that his company processing power designed to perform
humans if the robot arm struck them. might expand into, like computers in certain physical tasks using motor-
A robot with that capability might homes and small businesses. driven appendages and platforms.
need three-phase 240 volts AC at 50 The same applies to IBM Some computer-related devices such
amps or more — not an amount of Chairman, Thomas Watson’s remark as a laser or inkjet printer are as
power that you could obtain from a back in 1943: “I think there is a world complex as many robots since they use
battery pack in a mobile robot. With market for maybe five computers.” several motors to move the paper and
certain limitations in personal home As educated and savvy businessmen very accurately move the print head.
mobile robot design such as power who developed thriving computer However, they aren’t designed to
availability, size, and weight, designers companies, these weren’t flippant physically help us.
have strict boundaries within which to remarks made by people who didn’t
work. Fortunately, home and personal understand the computer business.
robots do not require great speeds of They were too focused on a narrow
Applications are the
operation which use a lot of power. market. There was no way that Key to Home Robot
There’s another major factor that Watson could know that in 60 years,
can make or break how well a robot there would be billions of computers Popularity
might sell in today’s cut-throat market: in the world — each ten thousand When the first home PCs found
74 SERVO 03/04.2019
their way onto our desktops, later brought parallel processing
there were very few applications power to small microcontroller-
available to the user. Gates and powered projects — especially
his company quickly realized to experimenter’s robots that
the need for an operating needed more channels of sensor
system when they saw the MITS input and motor control output.
Altair 8800 in 1975 and wrote The Propeller remains very
software for the computer using popular in robot design.
the “new” BASIC language.
The co-founder of that early Where to Start to
computer software company,
Micro Soft (originally spelled as Design and Build
two words) is still pretty well a Home/Service
known these days and is shown
in Figure 9. Figure 9 - Bill Gates with the Altair 8800. Robot
Figure 10 shows the CPU I started out this article
on top of a large floppy drive console MEMS motion sensors; intelligent discussing how service robots are
below. Do you remember those cameras; LiDAR, RADAR, and GPS becoming commonplace in our
things? People were amazed that MITS navigational sensors (to name a few), society. I then pointed out a few
sold the Altair computer kit with the higher power processing became examples of some home and social
microprocessor chip included for little necessary. Add to that processors for robot companies that seemed to be so
more than Intel’s 8080 chip alone sold scene analysis, speech recognition successful in the beginning but soon
for the previous year. and control, and what we like to call crashed for one reason or another.
The earliest home robots that artificial intelligence. Well, much more Does this signal that we are headed
went beyond basic relays and switches powerful processors had to come to to a robot recession? Absolutely
were forced to use programming the rescue. not! There are hundreds of service
languages and operating systems The Parallax eight-core Propeller robot companies that are producing
intended for human- thousands of new
based information robots of all types.
technology. Just as Yes, some will
the BASIC language immediately fail and
was adequate some a few years
for the early PCs later, and others will
with their eight-bit be lucky enough to
microprocessors, catch on with the
small DIY robots general public. Some
could easily operate of these companies
on the simple BASIC will be so successful
Stamp ‘micro boards” that others will
that today still use attempt to copy
the popular Parallax them to ride on their
PBASIC language coattails. This has
interpreter in its always been the case
Microchip PIC16F57- in any business, and
based microcontroller. iRobot soon found
When many companies
experimenter’s robots making copies of
began to take on their Roomba. A few
more sensors such were pretty good, yet
as: ultrasonic and IR most were very bad
object and distance copies. It has been
detectors; shaft said that imitation is
encoders; the more the most sincere form
complex series of Figure 10 - Altair 8800 computer atop floppy drive. of flattery, but not if
SERVO 03/04.2019 75
Figure 12 - Dr. Robot X80SVP
Wi-Fi mobile development
platform (pro version).

Figure 11 - Parallax Arlo robot platform.

that flattery is stealing your sells for $399 and is made


“cake.” by SPD. These robots allow
If you’re intending to a prospective home robot
design, build, and market builder to explore the newer
any style of service or home sensor technologies (such
robot, the best place to start as LIDAR) on an affordable
with is to do your homework budget.
on your basic idea. That other
95% of effort counts far more
than a sparkling new idea. Final Thoughts
Look at what’s on the market There is a beautiful
and analyze how well it is sunrise behind some of
selling — or not. See what the dark clouds that I have
features the robot has or does mentioned in this article
not have and if those features about home robots. Just as
prove to be necessary. Figure 13 - The SPD Mini RPlidar experimental two-wheel the fickle consumer buyers
What will certain drive platform. jumped on smartphones,
capabilities cost to build? SUVs, flat screen TVs, and
Decide if you are actually going to at 25 pounds, 18” diameter, and 10” Keurig-type coffee makers, the right
solve a need with your design. Do high — is equipped with object and home service robot will be developed
you really need the latest AI capability distance sensors and powerful motors and be snapped up by savvy buyers.
that seems to be the latest rage driving pneumatic wheels, and offers a The key is the ‘right’ robot.
incorporated in your robot? It has high quality test platform. Unlike those goofy-looking
been said: “Despite the latest studies The X80SVP Wi-Fi Mobile Cabbage Patch Dolls that were a hit
on AI, robots still have no common Development Platform (Pro Version) several years back (and now only have
sense.” Think about that statement. shown in Figure 12 costs $4,660 and value for extreme collectors), the right
Before deciding to begin cutting is made by the Canadian company, home robot will be a constant helper
metal to build a prototype, why not Dr. Robot. This is another smaller test in our homes. One of you readers just
buy an already-made kit or assembled platform that can be modified by an might be the inspiration, designer,
test platform like the Parallax Arlo experimenter. The even smaller Mini and builder of that electro-mechanical
shown in Figure 11 for $999. This RPlidar Experimental two-wheel drive helper that will place your name in the
robust robot platform — weighing in robot platform shown in Figure 13 history books. SV
76 SERVO 03/04.2019
GearBox setting your ideas into

MOTION
Use Small Mechanical
Components from
SDP/SI

START CREATING!

www.sdp-si.com 516-328-3300

Get Your Robot Runnin’!


Servo_Ad_5_24_18.indd 1 5/24/18 12:00 PM

SERVO 03/04.2019 77
FIRST AUTONOMOUS
DRIVING SHUTTLE BUS
FOR ALL WEATHER
CONDITIONS
Before GACHA,
Autonomous Cars
Operated Only in Ideal
Weather Conditions
... Now RoboSense’s
LiDAR “Sees” a 3D
World — Even in
Finland’s Snow and
Extreme Temperatures
Reaching -30°C (-22°F)

GACHA self-driving shuttle bus robo-taxi powered


with RoboSense all-weather LiDAR.
Environment Perception is Key in
Extreme Weather Conditions

A
s part of a collaboration, RoboSense (www.
robosense.ai) has designed a cold-resistant all- Environment perception is a huge task facing
weather LiDAR for the world’s first autonomous autonomous vehicle providers who require operation in all
driving shuttle bus for all weather conditions: weather conditions. Under extreme weather conditions,
GACHA. GACHA is equipped with RoboSense’s advanced the performance of critical sensors such as LiDARs and
cold-resistant 16-beam mechanical LiDAR environment cameras are severely impacted. Low temperatures such as
perception system to operate vehicles in harsh winter and those in Finland — where winter temperatures can reach as
other severe weather conditions. low as -30°C (-22°F) — can incapacitate the performance of
The autonomous shuttle bus robo-taxi, GACHA was semiconductor components, while snow- and ice-covered
designed in collaboration with Finnish autonomous driving roads essentially cripple the image algorithm’s ability to
company Sensible 4, who provided software for positioning, determine abstract environment information.
navigation, and obstacle detection; the Japanese company, In addition, on an extremely cold day, the motor of a
MUJI who provided expertise in design and user experience; multi-layered LiDAR might not even start up. It could also
and the Chinese company, RoboSense who provided the cause detection accuracy and ranging performance to be
advanced cold resistant LiDAR. diminished. Under these circumstances, RoboSense’s new
cold-resistant LiDAR steps in. It “sees” a 3D world through
Photos courtesy of Justus Hirvi/Bonzu & MUJI. emitting and receiving laser pulses, with point cloud
78 SERVO 03/04.2019
IN THE NEWS
algorithms that accurately
recognize obstacles — even GACHA at premiere with RoboSense
in snow and ice. According all-weather LiDAR.
to RoboSense, LiDAR is
the most important sensor
when it comes to bringing
autonomous driving
technology to extremely
cold environments.
“When it comes
to the GACHA project,
engineers from both sides,
through close and efficient
cooperation, made it
possible to let the whole
LiDAR system operate
steadily in the bitterly code weather in Finland,” said Mark in Finland and abroad.
Qiu, COO, RoboSense. “I’m pleased that our product has helped Sensible 4
and MUJI’s jointly designed GACHA self-driving shuttle bus
GACHA Robot Bus Fleet Deployed in solve perception problems in all weather conditions. In the
near future, all RoboSense multi-layer LiDAR products will be
Finland and Other Countries able to operate in automotive grade low temperatures under
RoboSense joined Sensible 4 on their first GACHA bus -40°C,” Mark Qiu explained.
fleet deployment this last April for use on public roads, and “The RoboSense LiDAR hardware and environment
will continue to participate in future projects in various cities perception algorithms we use are reliable and stable, even
at freezing temperatures of tens of Celcius degrees below
THE OWNERSHIP, MANAGEMENT, AND CIRCULATION STATEMENT OF SERVO zero,” commented Harri Santamala, CEO of Sensible 4.
MAGAZINE, Publication Number: 1546-0592 is now published bi-monthly, effective “The GACHA development got started when the
with the May-June 2018 issue. Subscription price is $26.95. 7. The complete
mailing address of known office of Publication is T&L Publications, Inc., 2279 Eagle Sensible 4 team — working back then with the first
Glen Pkwy, #112-481, Corona, Riverside County, CA 92883. Contact Person: Larry
Lemieux. Telephone: (951) 371-8497. 8. Complete Mailing address of Headquarters generation of robot buses — noticed that they just don’t
or General Business Office of Publisher is T&L Publications, Inc., 2279 Eagle Glen perform at all, even in light rain, not to mention the typical
Pkwy, #112-481, Corona, Riverside County, CA 92883. 9. The names and addresses
of the Publisher, and Associate Publisher are: Publisher, Larry Lemieux, 2279 Eagle winter conditions in Finland. Completely autonomous self-
Glen Pkwy, #112-481, Corona, Riverside County, CA 92883; Associate Publisher, driving technology is not here yet. Most self-driving cars can
Robin Lemieux, 2279 Eagle Glen Pkwy, #112-481, Corona, Riverside County, CA
92883. 10. The names and addresses of stockholders holding one percent or more operate only in ideal weather conditions and well-marked
of the total amount of stock are: John Lemieux, 2279 Eagle Glen Pkwy, #112-481,
Corona, Riverside County, CA 92883; Lawrence Lemieux, 2279 Eagle Glen Pkwy, roads. This is what Sensible 4 has managed to change
#112-481, Corona, Riverside County, CA 92883; Audrey Lemieux, 2279 Eagle Glen through repeated tests in the harsh winter conditions of the
Pkwy, #112-481, Corona, Riverside County, CA 92883. 11. Known Bondholders,
Mortgagees, and other security holders: None. 12. Tax Status: Has not changed Finnish Lapland,” Santamala continued.
during preceding 12 months. 13. Publication Title: SERVO Magazine 14. Issue Date Since its inception, RoboSense has been providing
for Circulation Data: October 2017-September 2018. 15. The average number of
copies of each issue during the proceeding twelve months is: A) Total number complete hardware and software LiDAR environment
of copies printed (net press run); 7,388 B) Paid/Requested Circulation (1) Mailed
Outside County subscriptions: 2,489 (2) Mailed In-County subscriptions: 0 (3) Paid perception solutions for the autonomous driving industry,
Distribution Outside the Mail including Sales through dealers and carriers, street solving the LiDAR perception challenges for autopilot
vendor, and counter sales and other paid distribution outside USPS: 2,209 (4) Paid
Distribution by other classes of mail through the USPS: 0; C) Total Paid Distribution: applications in all scenarios by combining the RS-LiDAR
4,698; D) Free or Nominal Rate Distribution by mail and outside the mail (1) Free or hardware and their
Nominal Rate Outside-County Copies: 0 (2) Free or Nominal Rate In-County Copies:
0 (3) Free or Nominal Rate Copies Mailed at other classes through the USPS: 0 (4) proprietary RS-LiDAR-
Free or Nominal Rate Distribution Outside the mail: 508; E) Total Free or Nominal
Rate Distribution: 508; F) Total Distribution: 5,206; G) Copies not distributed: 2,179 Algorithms. SV
H) Total: 7,385; Percent paid circulation: 90.24%. Actual number of copies of the
single issue published nearest the filing date is September 2018; A) Total number
of copies printed (net press run) 9,151; B) Paid/Requested Circulation (1) Mailed
Outside County subscriptions: 3,405 (2) Mailed In-County subscriptions: 0 (3) Paid
Distribution Outside the Mail including Sales through dealers and carriers, street
vendor, and counter sales and other paid distribution outside USPS: 2,471 (4) Paid
Distribution by other classes of mail through the USPS: 0; C) Total Paid Distribution:
5,876; D) Free or Nominal Rate Distribution by mail and outside the mail (1) Free or RoboSense all-
Nominal Rate Outside-County Copies: 0 (2) Free or Nominal Rate In-County Copies:
0 (3) Free or Nominal Rate Copies Mailed at other classes through the USPS: 0 (4) weather LiDAR on
Free or Nominal Rate Distribution Outside the mail: 500; E) Total Free or Nominal GACHA self-driving
Rate Distribution: 500; F) Total Distribution: 6,376; G) Copies not distributed: 2,800; shuttle bus.
H) Total: 9,176; Percent paid circulation: 92.16%. I certify that these statements are
correct and complete. Lawrence Lemieux, Publisher – 04/25/2019.
SERVO 03/04.2019 79
GIVING ROBOTS A
BETTER FEEL FOR
OBJECT MANIPULATION
Model improves a robot’s
ability to mold materials into
shapes and interact with
liquids and solid objects.

A new “particle simulator” developed by MIT


researchers improves robot’s abilities to mold
materials into simulated target shapes and interact
with solid objects and liquids. This could give robots
a refined touch for industrial applications or for
personal robotics.

A
new learning system developed by MIT researchers how liquids — as well as rigid and deformable materials —
improves robot’s abilities to mold materials will react to the force of its touch. As the robot handles the
into target shapes and make predictions about objects, the model also helps to further refine the robot’s
interacting with solid objects and liquids. The system control.
— known as a learning-based particle simulator — could give In experiments, a robotic hand with two fingers called
industrial robots a more refined touch, plus it may have “RiceGrip” accurately shaped a deformable foam to a
fun applications in personal robotics such as modelling clay desired configuration (such as a “T” shape) that serves as a
shapes or rolling sticky rice for sushi. proxy for sushi rice. In short, the researcher’s model serves
In robotic planning, physical simulators are models that as a type of “intuitive physics” brain that robots can leverage
capture how different materials respond to force. Robots to reconstruct three-dimensional objects somewhat similarly
are “trained” using the models to predict the outcomes of to how humans do.
their interactions with objects, such as pushing a solid box or “Humans have an intuitive physics model in our heads,
poking deformable clay. where we can imagine how an object will behave if we push
Traditional learning-based simulators mainly focus or squeeze it. Based on this intuitive model, humans can
on rigid objects and are unable to handle fluids or softer accomplish amazing manipulation tasks that are far beyond
items. Some more accurate physics-based simulators can the reach of current robots,” says first author Yunzhu Li,
handle diverse materials, but rely heavily on approximation a graduate student in the Computer Science and Artificial
techniques that introduce errors when robots interact with Intelligence Laboratory (CSAIL). “We want to build this type
objects in the real world. of intuitive model for robots to enable them to do what
In a paper that was presented at the International humans can do.”
Conference on Learning Representations in May, the “When children are five months old, they already have
researchers described a new model that learns to capture different expectations for solids and liquids,” adds co-author
how small portions of different materials — “particles” Jiajun Wu, also a CSAIL graduate student. “That’s something
— interact when they’re poked and prodded. The model we know at an early age, so maybe that’s something we
directly learns from data in cases where the underlying should try to model for robots.”
physics of the movements are uncertain or unknown. Joining Li and Wu on the paper are: Russ Tedrake, a
Robots can then use the model as a guide to predict CSAIL researcher and a professor in the Department of
80 SERVO 03/04.2019
IN THE NEWS
Electrical Engineering and Computer Science (EECS); Joshua other. The graph must learn to predict where and how
Tenenbaum, a professor in the Department of Brain and much all affected particles move, which is computationally
Cognitive Sciences; and Antonio Torralba, a professor in complex.
EECS and director of the MIT-IBM Watson AI Lab.
Shaping and Adapting
Dynamic Graphs In their paper, the researchers demonstrate the model
A key innovation behind the model — called the “particle by tasking the two-fingered RiceGrip robot with clamping
interaction network” (DPI-Nets) — was creating dynamic target shapes out of deformable foam. The robot first uses a
interaction graphs which consist of thousands of nodes and depth-sensing camera and object-recognition techniques to
edges that can capture complex behaviors of the so-called identify the foam. The researchers randomly select particles
particles. inside the perceived shape to initialize the position of the
In the graphs, each node represents a particle. particles. Then, the model adds edges between particles and
Neighboring nodes are connected with each other using reconstructs the foam into a dynamic graph customized for
directed edges, which represent the interaction passing deformable materials.
from one particle to the other. In the simulator, particles are Because of the learned simulations, the robot already
hundreds of small spheres combined to make up some liquid has a good idea of how each touch (given a certain amount
or a deformable object. of force) will affect each of the particles in the graph. As the
The graphs are constructed as the basis for a machine- robot starts indenting the foam, it iteratively matches the
learning system called a graph neural network. In training, real world position of the particles to the targeted position
the model (over time) learns how particles in different of the particles. Whenever the particles don’t align, it sends
materials react and reshape. It does so by implicitly an error signal to the model. That signal tweaks the model
calculating various properties for each particle — such as its to better match the real world physics of the material.
mass and elasticity — to predict if and where the particle will Next, the researchers aim to improve the model to help
move in the graph when perturbed. robots better predict interactions with partially observable
The model then leverages a “propagation” technique, scenarios, such as knowing how a pile of boxes will move
which instantaneously spreads a signal throughout the when pushed — even if only the boxes at the surface are
graph. The researchers customized the technique for each visible and most of the other boxes are hidden.
type of material — rigid, deformable, and liquid — to shoot a The researchers are also exploring ways to combine the
signal that predicts particle positions at certain incremental model with an end-to-end perception module by operating
time steps. At each step, it moves and reconnects particles, directly on images. This will be a joint project with Dan
if needed. Yamins’s group; Yamin recently completed his postdoc at
For example, if a solid box is pushed, perturbed particles MIT and is now an assistant professor at Stanford University.
will be moved forward. Because all particles inside the box “You’re dealing with these cases all the time where
are rigidly connected with each other, every other particle there’s only partial information,” Wu says. “We’re extending
in the object moves the same calculated distance, rotation, our model to learn the dynamics of all particles, while only
and any other dimension. Particle connections remain intact seeing a small portion.” SV
and the box moves as a single unit. However, if an area of
deformable foam is indented, the effect will be different.
Perturbed particles move forward a lot; surrounding particles
move forward only slightly; and particles farther away won’t
To post comments on this article and find any associated
move at all. files and/or downloads, go to www.servomagazine.com/
With liquids being sloshed around in a cup, particles magazine/issue/2019/02.
may completely jump from one end of the graph to the

Actuonix Motion Devices . ......41 Pololu . .......................................84


All Electronics Corp. ................77 Robot Power . ...........................55
ADVERTISER ExpressPCB ...............................7
Hitec RCD USA, Inc. .................2, 77
Routakit .....................................19
SDP/SI .......................................77
INDEX LDG Electronics ........................3 ServoCity ...................................77, 83
M.E. Labs . .................................49 Technologic Systems ...............45
PanaVise ....................................55 Tormach ....................................13
SERVO 03/04.2019 81
AVAILABLE AT goBILDA.COM
sales@goBILDA.com • 620.705.5959
Robots!

Clockwise from top:


Balboa, Zumo for Arduino,
Romi, Zumo 32U4, 3pi,
Balboa (w/stability conversion kit),
Romi (w/robot arm kit)

Find these robots and more at www.pololu.com

You might also like