International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395 -0056

Volume: 03 Issue: 09 | Sep-2016 www.irjet.net p-ISSN: 2395-0072

Tesla Autopilot : Semi Autonomous Driving, an Uptick for Future

Autonomy
Shantanu Ingle1, Madhuri Phute2
1Department of E&TC, Pune Institute of Computer Technology, Pune, Maharashtra, India

2 Department of E&TC, Pune Institute of Computer Technology, Pune, Maharashtra, India

---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract - Manual driving has certainly changed the highway(freeway), change lanes and adjust speed in
way we travel and explore different parts of the globe, with response to traffic conditions. Once the driver arrives at
total human control and with human intuitions and the destination, Model S or Model X scans for a parking
previous driving experiences. But manual control doesn’t space and parallel parks on the driver's command[2]. In
ensure safety and zero fatalities in every possible driving the new Autopilot, the instrument cluster’s new driver
condition. Autonomous driving is the answer to that focused design shows the real-time information used by
problem. However, artificial intelligence hasn't yet touched the car to intelligently determine the vehicle’s behavior in
every aspect of technology. In this paper we present a brief that moment relative to its surroundings[3]. Along with
overview about how Tesla Motor’s semi autonomous the autopilot feature, the driving behavior of each Tesla
driving technology called Tesla Autopilot is changing the car, while navigating through different road conditions, is
way we navigate and transport by incorporating state of shared to its central server. Based on machine learning
the art current artificial intelligence and hardware and Tesla design engineer recommendations, a new
feature update is developed and launched to every other
technology, and enhancing the next generation
Tesla car on planet. The Instrument panel provides a
autonomous driving with real-time driving updates. Tesla
visualization of the road as detected by the car’s sensors,
Autopilot relieves drivers of the most tedious and
giving drivers the information their car is using for
potentially dangerous aspects of road travel.
features including lane departure, blind spot detection,
Key Words: Tesla Motors, Autopilot, Electric Vehicles, speed assist, collision warning, adaptive cruise and auto
Radar, Ultrasonic sensors, Autosteer, Autolane steer. Based on previous research done on self driving
cars(not yet commercialized) and hardware resources
change, Autopark
available to consumer market, which enables Tesla’s
1. INTRODUCTION autopilot feature, we propose our views and assumptions
of the operation and working of individual autopilot
Tesla claims that Autopilot gives you more confidence hardware modules.
behind the wheel, increases your safety on the road, and
makes highway driving more enjoyable. While truly 2. MODULES OF AUTOPILOT
driverless cars are still a few years away, Tesla Autopilot
functions a lot like the systems that airplane pilots use 2.1 Radar and Camera Combination:
when conditions are clear. The driver is still responsible
The front bumpers of Tesla Model S and Model X have
for, and ultimately in control of, the car[1]. What's more,
radars behind them, with a range of several hundred
Tesla gives the driver an intuitive access to the
meters, that can detect cars and moving objects from a
information which it is using to control its actions. Along
substantial distance. However, unfortunately, the radar is
with the usual combination of accident prevention
unable to detect lanes or motionless objects like
technology such as Advanced Driver Assistance
stationary humans.
Systems(ADAS), which actuates emergency steering and
braking, the autopilot technology that powers the Tesla
Model S and Model X electric vehicles enables cars to
autonomously steer, change lanes, follow vehicles and
curves, and park automatically in the garage. The nature
of these cars is boldly different from most of the other
production vehicles out there in the consumer market.
Tesla introduced its Model S software version 7.0 which is
a software update for Tesla’s autopilot hardware for
Model S and Model X production vehicles, which allows
vehicles to use data from the surrounding cameras, radar,
and ultrasonic sensors to automatically steer down the

© 2016, IRJET | Impact Factor value: 4.45 | ISO 9001:2008 Certified Journal | Page
369
 International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395 -0056
Volume: 03 Issue: 09 | Sep-2016 www.irjet.net p-ISSN: 2395-0072

Figure 1: Vehicle and Pedestrian Detection

A camera system is an average wide angled camera

equipped on the front or in a surround configuration on
the car roof which recognizes various objects such as cars,
trailer, bike, cyclists , pedestrians and road markings.

Figure 3: Road Type and Curve Detection 2.2 Vehicle

Tracking with Ultrasonic Sensors

Ultrasonic sensors, which are mounted on corners of the

car body, detect vehicles on the adjacent lane during lane
changing or auto-steering. These ultrasonic sensors are
already in cars that offer automatic ‘Reverse Park Assist’
technology, which help navigate the car into tight reverse
Figure 2: 3-D Mapping of Each Individual Vehicle parking spots.

The chip behind autopilot on Tesla is Mobileye Eye Q3

processor which is based on the 40nm process. The tiny
silicon packs 4 cores clocked at 500 Mhz with 64 MACs
per core. The Eye Q3 processes the inputs taken from
satellite imagery, radars, ultrasonic sensors and camera,
and actuates steering and speed control systems.
Mobileye has deployed the first DNN (Digital Neural
Network) on the road with Tesla electric vehicles and is
responsible for 4 major jobs; Free Space Pixel Labeling,
Holistic Path Planning, General Object Detection and Sign
Detection.

The DNN was trained with various side, front and rear
sides of various cars until it is able to detect them to a
reasonable accuracy and consequently construct a 3D Figure 4: Blue car is in the blind spot of red car’s
model of the same. Free Space Pixel Labeling is basically driver. The blue arcs depict the sensor range of the
recognizing the area on-camera which is obstruction free. red car’s sensors. The orange rectangles depict the
It is also the area upon which the car will be allowed to go. critical zone which should be supervised.
Holistic Path Planning capability of the Mobileye chip
which allows it to decide the way forward with very little The Tesla car is equipped with an array of 12 ultrasonic
visual cues. This is the process which tells the car where sensors. The single sensors are placed equally on its front
to drive and also controls the steering. The chip on-board and rear. For the job in the above diagram only six of them
the Tesla is capable of identifying over 250 signs in more are used, three on each side of the car: the front-side and
than 50 countries[4]. These include everything from turn the rear sensor (aperture 75 deg.) and additionally the
signs to speed limits. The system is also capable of passive rear-side sensor (aperture 50 deg.), which only
identifying and interpreting traffic lights, road markings receives ultrasonic echoes emitted by the rear sensor. The
and general items such as traffic cones. The system has particle filter with mixture tracking capabilities[5] to
the capability to detect the road surface as well as any perform Bayesian filtering in terms of Monte Carlo
debris present. This allows the Tesla car to be aware of sampling is implemented to fuse signals from all
not only what kind of road the car is traveling on ultrasonic sensors[6].
(highway, country side etc) but also to detect debris and
other undesirables such as potholes on the road (and
consequently avoid them).

© 2016, IRJET | Impact Factor value: 4.45 | ISO 9001:2008 Certified Journal | Page
370
 International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395 -0056
Volume: 03 Issue: 09 | Sep-2016 www.irjet.net p-ISSN: 2395-0072

when AutoLane Change is engaged is simple; engage the

turn signal and Model S will move itself to the adjacent
lane when it’s safe to do so[8].

Figure 5: Coverage of the used ultrasonic sensors

(black arcs) and blind spot zone. The coordinate
system is fixed to the host car as shown and is the
base system for the particle filter.

2.3 Satellite Imagery Figure 7: AutoSteer & AutoLane Change 3.2 Automatic
Emergency Steering and Side Collision Warning
Road monitoring using satellite system consists of three
parts: an in-car unit, a central computer server and a Side collision warning further enhances the active safety
mobile satellite communication system. The in-car system capabilities of Model S by sensing range and alerting
reads the vehicle's position every second using the Global drivers to objects such as cars that are too close to its side.
Positioning System (GPS). An algorithm called the ‘map When the car detects an object close to its side, fluid lines
matcher’ uses that position and tachometer data, which will radiate from the Model S image in the instrument
has vehicle’s speed, to identify the road on which it is panel to alert the driver.
being driven. Any traffic congestion is automatically
detected for each road segment, based on a prior
knowledge of that road, such as the expected speed of
traffic under non- congested conditions.

Figure 8: Blindspot detection and side collision

warning while lane changing and steering 3.3
Figure 6: Road Mapping and Vehicle Tracking AutoPark

3. TESLA AUTOPILOT FEATURES Model S can parallel park itself, eliminating the need for
drivers to worry about complex and difficult parking
3.1 AutoSteer and AutoLane Change maneuvers. When driving at low speeds around cities, a
“P” will appear on the instrument panel when the Tesla
AutoSteer keeps the car in the current lane and engages car detects a parking spot. The Auto-Park guide will
Traffic-Aware cruise control to maintain the car’s speed. appear on the touch-screen along with the rear camera
Using a variety of measures including steering angle, display, and, once activated, Auto-Park will begin to park
steering rate, and speed to determine the appropriate itself to by controlling the steering and the vehicle
operation, AutoSteer assists the driver on the road, speed[9].
making the driving experience easier[7]. Changing lanes
© 2016, IRJET | Impact Factor value: 4.45 | ISO 9001:2008 Certified Journal | Page
371
 International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395 -0056
Volume: 03 Issue: 09 | Sep-2016 www.irjet.net p-ISSN: 2395-0072

With the new update of "Summon", once you arrive home proven by talking on concrete grounds, and looking at the
and exit Model S or Model X, you can prompt it to do the exponential increase in processing capabilities, and
rest: open your garage door, enter your garage, park itself, progress being made in leaps and bounds especially by
and shut down. In the morning, you wake up, walk out the Mobileye and Tesla Engineers, that it looks like the first
front door, and summon your car, it will open the garage hints of fully autonomous driving might be here by 2018.
door and come to greet you. More broadly, Summon also
eliminates the burden of having to squeeze in and out of We are finally seeing the vestiges of something resembling
tight parking spots. During beta stage of Summon, it the promised future, where cars are fully autonomous and
expects customers to become familiar with it on private manual driving is frowned upon.
property. Eventually, the Tesla car will be able to drive
anywhere across the country to meet the customers, 5. REFERENCES
charging itself along the way. It will sync with customer's
calendar to know exactly when to arrive[10]. [1] Tesla Model S Wikipedia:
https://en.wikipedia.org/wiki/Tesla_Model_S.
[2] Tesla Motors Website:
https://www.tesla.com/models. [3]
http://teslatap.com/articles/instrument-panel-7-
0user-interface-evolution-and-analysis/
[4] S.A. Larsen H., Koren, and Rune Solberg, “Traffic
monitoring using very high resolution satellite
imagery,” Photogrammet- ric Engineering & Remote
Sensing, vol. 75, pp. 859–869, July 2009
[5] J. Vermaak, A. Doucet, and P. Perez, “Maintaining
multimodality through mixture tracking,” in Proc.
Ninth IEEE Int Computer Vision Conf, Nice, France, Oct.
2003, pp. 1110–1116.
[6] A. Doucet, N. Defreitas, and N. Gordon, Sequential
Monte Carlo Methods in Practice, "Statistics for
Figure 9: AutoPark - Summon Engineering and Information Science", 1st ed. Springer
New York, June 2001. [7] Tesla
4. CONCLUSION Website:https://www.tesla.com/videos/
autopilotautosteer [8] Tesla
Tesla Motors warns that Tesla Model S and Model X Website:https://www.tesla.com/videos/autopilot-
requires drivers to remain engaged and aware when lanechange [9] Tesla
AutoSteer is enabled. Drivers must keep their hands on Website:https://www.tesla.com/videos/
the steering wheel, as they claim it hasn’t yet reached full autopilotautopark
autonomy. Consumer-report experts believe that these [10] https://www.wired.com/2016/03/teslas-
two messages— your vehicle can drive itself, but you may
summonfeature-like-knight-rider-kinda/
need to take over the controls at a moment’s notice— [11] http://jalopnik.com/does-teslas-semi-
create potential for driver confusion. Following a series of autonomousdriving-system-suffer-from-
crashes, one of which - the Florida crash - was fatal, has 1782935594 [12]
prompted investigations by National Highway http://www.wsj.com/articles/mobileye-
Transportation and Safety Association [NHTSA] and the endspartnership-with-tesla-1469544028
National Transportation Safety Board [NTSB][11]. As of
now, Tesla claimed the crash occurred due to inefficiency
of the Mobileye Q3processor, to which Mobileye ended
their collaboration with Tesla motors, but they will
continue providing technical assistance to Autopilot with
no further development[12].

Now as a footstep for achieving complete autonomy, Tesla

is constantly introducing enhancements in its software
and hardware, proven over millions of miles of internal
testing, to ensure that drivers supported by Autopilot
remain safer than those operating without assistance. To
be fair to humans however, there are various scenarios in
which an AI powered car would be completely incapable
to respond but a human would be able to. But results have
© 2016, IRJET | Impact Factor value: 4.45 | ISO 9001:2008 Certified Journal | Page
372