BRAIDED PNEUMATIC MUSCLE ACTUATORS

Darwin G. CaldweU, A. Razak and M. Goodwin

Dept. of Electronic Eng .• University of Salford. Manchuter. MS. 4WT. UK.

Abstract. This paper considers the requirements for a flexible actuation system that can be applied to
untethered robots. Using organic muscle specifications as a base, mathematical models of the drive forces
and operating principles in a braided pneumatic actuator are studied. Experimental testing of the braided
muscle will sbow that it has the traditional benefits of pneumatic cylinders, combined with power/weight
outputs of over l.SkW/kg at 200kPa. In addition, reductions in the sponginess associated with air have
made position control possible. This potential for controlled motion is demonstrated on a jointed fmger
powered by 2 pairs of antagonistic muscle.

Key Words. Actuators; robots; pneumatic power systems; artificial limbs; modeling

1. INTRODUcnON
There has also been renewed interest in hydraulic
The perfonnance of actuators is critical in detennining systems prompted by developments in Electro-
the strength. speed of motion., power, response rate, Rheological Fluids (Scotl. 1985; Block. 1988) which
compactness and in many cases the compliance can reversibly change viscosity from liquid to 'solid'.
exhibited by robots. Unfortunately present drive Use of this technology has lead to the development of
systems (usually electric, hydraulic or pneumatic), prototype actuators and compact valves with rapid
seldom achieve these demands due to well known and switching characteristics.
well documented drawbacks that place constraints on
their application in future generations of autonomous or Researchers have also developed actuators based on
semi-autonomous robots (McKerrow, 1991; Caldwell, organic long chain polymers (Katchalsky, 1949;
1990a). To overcome these limitations research efforts Caldwell, 1990b; Wassermann, 1960). These polymeric
have been directed towards a number of new fonus of (pseudo-muscular) actuators are chemical based devices
actuator technology. designed to take advantage of the hugh energy storage
potential of chemical substances. The mechanism can
Piezo~lectric actuators based on a variety of materials be triggered by a variety of techniques with the fastest
such as quartz, PZT (lead zirconate titan ate) or PMN method at present using water/acetone to give a
(lead magnesium niobate) have been successfully contractile rate of 10%/sec (CaIdwell, 1992). Although
developed. Typically these actuators are; compact, this response rate is poor compared to natural muscle
controllable with accuracy of a few pm. have fast the forces generated are comparable (up to 3ON/cm2),
response speeds (lmsec. or better), generate high forces and with thinner fibres (10 um) it is predicted that the
(440N/cm 2), and have a driving power a magnitude response rate could be increased by up to 100 times.
smaller than electro-magnetic motors (Takahashi, 1987).
The main applications of this technology are in There have also been attempts to improve the
micro-actuators and in micro-manipulation where good performance characteristics of pneumatic systems
resolution on shape changes of O.lmm or less is (Wolff, 1970; Bridgestone, 1988; Heoke, 1969). Several
possible. of these actuators operate by controlling the expansion
The use of the Shape Memory Effect has also aroused or contraction of pressurized units using flexible
considerable interest (Dario, 1987; OynaUoy, 1989). restraining fabrics. Using these techniques systems have
Shape Memory Alloys (commonly Nickel-Titanium, been reported that produce:
although other alloys are available) exhibit a peculiar 1. Accurate smooth motion from start to stop.
property called the shape memory effect which leads to 2. Low hysteresis and friction.
dimensional changes with variation of temperature. 3. Low cost, yet powerful actuation from a light
Using these alloys large recovery stresses (up to 500 weight compact device.
MPa) can be obtained within relatively small spaces, 4. Rotation as well as longitudinal motion using the
and this has been used as a power source for robotic same basic driving principle.
applications. At present the major drawbacks with these
Wlits are the relatively long cycle times, complex Pneumatic systems are of particular interest since they
control problems, low efficiency «10%) and long term satisfy many perfonnance requirements, having a
degradation of the SME. variety of advantages: Cheapest form of drive; Few

522
moving part 10 reliable with low maintenance; the ranges 15N-40N/cm2 and 25%/s to 1800%/s
CommOll technology; Quick DOll and response time; respectively.
Direct drive possible; Safe in explosive or wet 2. Good .olion control; repeatability and
environments; Compact; Simple control, bang-bang; lCCW'8Cy. Actuators must be capable of cantrolled
Intrinsic compliance, while failing primarily on the motion with sensory feedback of relevant data;
issue of control, which precludes their extensive use in position, velocity and tension. Muscles have in-built
robotics. The two major problems traditionally servo feedback giving accurate control. This internal
associated with the use of pneumatic devices muscle feedback (kinaesthetic data) is augmented by
(cylinders) are: data from the skin (cutaneous data) and the eyes giving
i) the accuracy and difficulty of cantrol. very accurate control.
ii) compliance (sponginess) which means that variations 3. Energy Storage. The actuators should be able
in loading effect the position. In effect both these to operate for extended periods from a self contained
problems are due to a common factor, the power supply unit. A typical period between meals
compressibility of the fluid (air), and although partial (muscle refuelling) is 4-6 hours, although by using
solutions have been found by carefully controlling the reserves this can be much longer. This is a major area
venting of the cylinders the problems largely remain of developmental concern in robotics, especially as
(Hu, 1991). future requirements will mean that more robots are free
roaming.
These activities all aim to improve actuator and overall 4. Robust and dependable. Actuators must be
robot performance. However, to produce the 'ideal' sufficiently rugged to withstand industrial environments
robotic actuator a basic specification is requirement. with specialised functions for extreme locations (space,
Fortunately, there is already an excellent model for a nuclear, sub-sea. extreme heat or cold). Muscle is
flexible, portable, general purpose robot actuator: envirorunentally protected by skin and can operate in
organic muscle. extremely hostile locations when provided with
protection. In addition it typically has a working life in
Using muscle performance has a base specification for excess of 65 years.
a robotic actuator system, this work considers the 5. Good efficiency. To conserve energy,
development a new braided pneumatic muscle actuator. efficiency should be as high as possible. Muscle
This design has the good characteristics of traditional efficiency for direct conversion from chemical to
pneumatic cylinders, combined with very high mechanical energy is generally in the range from 45%
power!weight ratios, significantly improved control up to 70% (Wilkie, 1976).
potential and reduced the system sponginess. The study 6. Environmentally Sare. Ideally the unit should
of the operation will include a theoretical analysis of produce no chemical or noise pollutants. Muscle
the forces involved in the power cycle, and the drive produces COl'
force capability, and experimental verification of these
concepts. While the factors cited above show the positive
The actuator performance will be demonstrated using features of natural muscles it is important that new
a three jointed finger with coupled pairs of muscle actuation systems should also retain the advantages of
(flexor and extensor) driving each joint. This will show conventional man-made systems:
the excellent actuation properties of this pneumatic i). Continuous prolonged operation without tiring or
muscle design combined with system controllability. performance degradation.
ii). Ease of replacement when irreparably damaged.
iii). Low cost and ease of maintenance.
2. ORGANIC MUSCLE AS AN ACTUATOR

Animals use 'muscle' (skeletal) to exert forces and 3. ACTUATOR DESIGN STRUCTURE
move objects including themselves. These muscles, are
biological machines of great complexity, which convert The braided pneumatic actuator is constructed as a
chemical energy (food) into forces and mechanical two-layered cylinder, fig. 1. The inner layer is made
work (Wilkie, 1976). The basic portability. efficiency from double lined thin walled rubber tubing which
and effectiveness of this natural system makes it an when capped by perspex end-plugs forms a pressurized
ideal base specification. contairunent unit. The outer layer is formed from
interwoven (but not bonded) nylon fibres producing a
By considering the performance of a range of human flexible structure with high tensile strength. This layer
and animal muscle systems the following criteria have prevents the rubber tubing from over-inflating and
been suggested as the fundamental performance rupturing, while acting as a platform for the
standards that will have to be achieved, or bettered by transmission of the pneumatic energy from the liner to
actuators to be used in future free-ranging robotic the driven joint.
systems. <;$" ~~("" ~-
1. Power/weight ratio. This ratio which should
include the mass of the energy source such as batteries J ./. {«?~
' ,~), ~
, . ?J
:, ... " >-..>;~ -
or hydraulic pumps is vital in determining the speed of
movement and strength of a robotic system. Typically /
Rubber I nne r
muscle operates in the range from 40W/kg up to L.yer Shell
225W/kg. but has the potential for peaks as high as Fig. 1. Braided Pneumatic Muscle Actuator Design
lkW/kg (Wilkie, 1976). The power/weight ratio for
muscle is derived from two factors; the contractile The two perspex plugs that form the termination
force and the rate of contraction which have values in connectors for the system act as the air inlet-outflow

523
channel and have a diameter slightly larger than the Variations in the interweave angle (within these limits)
unstJessed diameter of the rubber liner. The liner and effect the surface area, volume, intcmal pressure. and
the nylon shell are bonded to the end-caps using a the internal forces. These variations, and bow they act
flexible adhesive and clamped securely in place using to produce the actuator drive force will be studied in
three rubber sealing rings. In operation this wlit can the following sections.
withstand preBSlD'CS up to 450kPA (4.5 bar), although
200-300 kPa is more usual. The mass of the actuator
is 5.5g. 4.1 Surface Area v Interweave Angle
In its natural unstrained condition the dimensions of From experimental measurements of the structure when
the actuator and the interweave angle (angle of fibre fully compressed and fully stretched it was found the
cross-over), fig. 2 are: interweave angles were 68.So and 19.2" respectively.
Length = l40mm Diameter = 5.Omm Using this data and the measurements for the fully
Interweave angle = 22.3-. compressed length and diameter, and the fully stretched
length and diameter, the braided actuator can be
modelled as a flexible trapezoid, fig. 2.
Interweave
Angle In the quadrilateral 1--. is the measured length of the
actuator when fully stretched, and t- is the length
when fully compressed. Using this information the total
length of a single nylon fibre can be calculated using
either cm;npressive or expanded data:

h = (L..}2)/cos a... = (L,Jl)/cos a.. (2)

D • where a is the interweave angle, and h is the actual
Fig. 2. Interweave Angle physical length of a nylon strand. By using these twin
estimates for the fibre length h. the measurements
inaccuracies are minimized.
4. POWER GENERATION POTENTIAL
Calculation of the radial component of the fibres in
As a compressed fluid derived system, the basic the trapezoid can also be obtained where f represents
relationships governing the generated forces stem from the fibre circumference encirclement length. This gives
the standard Gas Laws and in particular (Tre Ioar,
1975); f = (D.J2)/sin a... = (D..J2)/sin a.. (3)

F= PA (1) where D ... is the diameter of the actuator when fully

elongated and D-. is the diameter when fully
where F is the force, P is the supply air pressure, and compressed.
A is the area over which the applied pressure acts.
Using this information the surface area at any
In a conventional cylinder the moveable surface on interweave angle can be determined;
which the internal pressure can act is the area of the S. = Ti * D * L (4a)
piston face-plate. This can be a relatively small
proportion of the pressurized area. By contrast, in the S. = '7r * (f sin a) * (h cos a) (4b)
braided muscle system the whole length of the shel1 is
a deformable structure on which the internal pressure Similarly. if the length or diameter of the actuator is
can act to generate a drive potential. known the interweave angle can be calculated:
a = cos'· (0.5· L)/h (5a)
In this pressurized, stressed, state the braided or
interwoven fibres are still free to move/distort and the a sin'· (0.5· D)/f = (5b)
structure can still be stretched or compressed with the
physical limits set by the width and length of the Using (4b) the variation in surface area with
nylon fibres, fig. 3. The elongation limits are: length = interweave angle can be plotted. This surface
143mm; diameter 4.8mm and interweave angle 19.2·, = area/interweave angle relationship has been normalised
while the compression limits are: length =
48mm, and is shown in fig. 4.
diameter = 14.2mm and interweave angle 68.8·. =
. ' ""<~
. .---- \
:'
C01llpr ...... d
.../ .' \,
\\
'

., I
• I

, f

Elongated Actuator .. ~ $" ' ~I "",ee',oj Ar{.:.?

'C 1':: 1;:' .:, ',( ec 70 f!" ; " ~' n:· 8Q '.g,j,C,;

Fig. 3. Compressed and Elongated Actuator Structure Fig.4. Surface AreaNol. v Interweave Angle

524
'Ibis analysis nweals that the maximum lW'face area parts:
is attaiDed when the in1erweave IDgle is 45·, giving a i) End -plates: the internal cylinder pesaure impinging
crossover mgle of 90". CD the end plates tries to force the shell to expand.
'Ibis forces always causes elongation as in a
conventional cylinder.
4.2 Volume v Interweave AnRle il) Braided nylon shell: This is by far the larger force
caused by pressure on the containment shell. This force
For conventional pneumatic cylinders the force output is bi-directional and always acts towards the natural
is dependent on the internal pressure and the area of minimum energy point
any moveable surfaces. In a cylinder this latter factor
is represented by the cross-sectional area of the vessel In the neutral minimum energy/force state the internal
ie the eod-plate area. In the pneumatic muscle the energy is given by:
drive pressure is applied to the whole surface skin of F,... = S. P + lE. P (8 a)
the containment shell. Under these circumstances
pressurization causes the muscle to take up a volume = p. (S. + 2E.) (8b)
which is maximized (minimum pressure), ensuring
minimum internal energy ie its most stable state. where E. is the area of the end plates.

From studies of Gas laws it is known that the When deflected from this starting point by stretching
relationship between pressure and volume is given by or compressed a force is generated that is represent
( ,1986); by;

(6) F = S. p' + 2E. P' (9a)

where n is the number of moles, R is the Gas = p. • (S. + 2E.) (9b)

Constant, and T is the absolute temperature.
where p. is the new internal pressure. This pressure
From the surface area calculations (2)-(5b) it has been will be determined by the volume changes as the
shown that the actuator length is L =
h cos a and the actuator elongates or contracts. The pressure change
diameter is D = f sin a. The volume of the cylindrical with interweave angle is the reciprocal of the volume
structure adopted by the muscle actuator is given by: change outline above.

Vol = (1)2/4) • 7t • L (7a) The driving force that can be used by the actuator is
or given by:
Vol = K (sin a)(sin a)(cos a) (7b) F. F - F ... (10) =
where K =(pi/4) • f · h. Plotting the nonnalised Where the force is positive the actuator will contract,
volumefmterweave angle relationship, fig. 4, it can be while a ~egative force will produce expansion.
seen that the maximum volume, and therefore the
minimum pressure and energy state occurs at an Figure 5 shows the nonnalised force output profile. At
interweave angle of 54°. the maximum volume state (interweave angle 54°). =
The internal force has been nonnalised to L Distortion
This is the mmlIDum force/energy state and the from this point will produce changes in the internal
actuator will always try to return to this position if pressure which fonns the driving force. Reduced
distorted. interweave angles (elongation) provide a positive force
in excess of I, while compression produces a value
Based on the estimate of the natural pressurized less than 1 and a negative pressure relative to the
interweave angle the length of the actuator when neutral point.
inflated was calculated as 89nun. This is in excellent 4 t1ormeli • • d Out gut

agreement with the measured value of 9Omm.

Increasing the pressure to its maximum test value of
450KPa does not have any effect on the length. :s ~ '"
.•. ••••• -- ~:- -------.-- .. •.• •-•••• .•.•.. .••••••• ...• -.-... .... . ... •_

4.3 Actuator Driving Force 2 '.

~ .• • • • - - - - - •• -.-. :.:. • • •• •. • • - .- - • ••• • • .•• •. •• • - • ••• • • .. . ..• • •.. . . .. -. -

""" .
-:. . .
Operation of the braided actuator as a drive system 1 ~ ___.. _.. . __ __.. _.. .. ...... ........ .. .. ____ .. __~:~ :.~.•~'V. !' ........ !.;: "~:'.: "; •. : ::~.~ """" """ "" "
relies on a critical understanding of the importance of
the minimum energy state (interweave angle 54°) and
the fact that any deflection from this point will produce .~ ____ ~~ ____ ~ ____ ~ ____ ~ ____ ~ ____ ~ ____ ~--J

• •• 28 :S8 .8 ~8 ~a 7a .a 9111
drive forces acting to return this state. In attempting to
return to this original minimum energy/force state
useable actuation forces are generated. Fig.5. Int~rnal Force v Interweave Angle

With conventional cylinder designs only the piston face This theoretical analysis assumes perfect conversion of
plate is free to move and hence a driving force is only pneumatic energy with no losses. Unfortunately, as
applied to this area. In the braided muscle the active with all systems there are losses. These arise from
pressurised area can be considered in two separate factors such as the elasticity of the rubber, expansion

525
energy for the liner, dead spaces during expansion, with a range of masses from O.5kg to Skg and the
friction between the inner and outer layers, and initial eloogatioo was measured. The relationship
leakages and pressure losses in valves, piping and the between the elongation and the load followed the
actuators. To take these effects into account an profile expected of a rubber spring system
efficiency or conversion factor can be included in (10). (Treloar,I97S). The isotonic results are shown in fig.
This efficiency rating has been obtained from 7.
experimental data.
The modified equation is: F4 = E.<F - Folio) (11) The relationship between contraction and pressure is
where Be is the force conversion factor. logarithmic and in good agreement with the theoretical
and simulated predictions (Caldwell, 1993). As with the
isometric tests the forces are lower than expected 57%,
5. SYSTEM TESTING giving a combined force conversion factor of 0.55.

Tests were conducted to determine the actuator

response under isotonic and isometric conditions and 5.3 PowerlWeight Relationship '
power/weight ratio.
To measure the power/weight output the actuator was
5.1 Force/Pressure Relationship aSOMETRI9 loaded as described for the isotonic tests. Motion of
the actuator was measured by a rotary position sensor
The forces generated during contraction of the muscle connected a PC used to monitor and display system the
were measured using a load ceIJ, with the pneumatic responses. The performance of the system was assessed
muscle connected to the load cell using Kevlar fibres. using loads of 1kg, 2kg and 3kg and operating at
pressures in the range 0-450kPa. The power/weight
The actuator was extended to its test length under zero performance under a series of step input driving signals
pressure and the load ceIJ output was zeroed. When was also plotted. fig. 8.
energised, the rubber liner inflated generating a
contractile force as the system contracts to regain its 14r----------------------------------
POUol_r (Watt_)

maximum volume state. Changes in the contractile 12

force were measured as the pressure was cycled from 18

OkPA to 450kPa and back to OkPa. fig. 6.
2. [.. . . 8
. ' 75%
,~ ............P. !?1}.~.~.!: . ~. !?1}................ ~ .............. .

....
%
~

.. ........ ... ...... ....... . . ····F. · ;:':.~ ... ........ , .. . . 6.0.%. ... .
2
u,
..
• 5 ·~~~i~::::~.--:-::-·:::-:~·-::=:2.5·;.·· .. 8~----~----~----~----~~----~~
8 4 ~

................... MJ.. .. eJc.p ..

·.~~~-----2~--~3~--~----~s 11<.. -.- ~I<.. . .• ...
• __ .... _ ... .a._.....
- - 25"
........ , ---..-6."
. ,,,, Fig.S. Power Output v Pressure
Fig.6. Isometric Tests
As would be expected power is closely related to the
These isometric (constant length) tests show an almost pressure and the relationship is basically linear. It was
linear relationship between pressure and the contractile observed that the power output is dependent on the
force, with the force being dependent on the initial load, with the maximum power being obtained with a
elongation of the system. These results agree with the loading of 2kg. Under these conditions the
theoretical and simulated profiles (CaldweIl, 1993), power/weight ratio of the braided muscle system was
although the forces are only 53% of their predicted l.SkW/kg at 200kPa (2 bar) and 3kW/kg at 400 kPa.
values due to system losses, ie the effective force
conversion factor. 6. MANIPULATOR SYSTEM CONTROLLER

These initial results indicate that this form of actuator

5.2 Displacement/Pressure Relationship aSOTONlC) has considerable power advantages when compared to
electric motors and conventional pneumatic cylinders.
However: an effective actuator must combine good
E •
u power/weight performance with controllability.
., 3
C
QJ
S
~2
III
....
0- Fig. 9. Finger Design
...o
V' 1

To provide a basic form of system test a simple 3

o~~~--~----~--~--~
0 2 3 jointed finger was constructed based on an
Fig. 7. Isotonic Tests ~'*,. -;8 (x 10OkPII) anatomically accurate model of the human fmger. As
in the human hand the second and third links are
With the muscle unpressurized the actuator was loaded coupled, fig 9. This test mechanism was chosen as the

526
power!weigbt and power/volume ratios, and damage and power/volume ratios. By improving the system
preventing compliance inherent to pneumatic designs is efficiency the possible outputs as predicted by
a coosiderable benefit in dextrous manipulator theoretical and simulation studies are very impressive.
applications. In addition pulsed control of this system using valves
with low electrical power requirements has shown the
& the poeumatic muscle actuator operates only in a control potential of this system using a jointed fmger
contractile mode, two actuators arranged as antagonistic as the base test system. Manipulators can particularly
pairs (extensor and flexor) power each joint The benefit from the high power and compliant control of
proximal and middle phalanges are directly coupled to this technique.
the actuators using Kevlar, while the distal phalange is
coupled to the middle joint, to form the natural primate Clearly the combination of advantages of traditional
curling action. pneumatic cylinders with controllable braided
pneumatic muscle actuators holds considerable
Although the system has 4 active actuators only 2 are potential.
used in controlling the position. These are the flexor
muscles for both the proximal ,and middle joints. These
actuators are controlled by a simple pulsed mode PO 8. REFERENCES
controller with a limited form of gain scheduling and
position feedback to the AID converter port of the PC. Block H. and Kelley J.P.(1988), "Electro-Rheology",
The extensor muscles are used in regulating the force Journal of Physics D: Applied Physics, Vol.21,
and system compliance. pp: 1661-77.
Bridgestone Corp. (1988), "The Rubberuator", Product
Actuator control is achieved by piezo.electric valves Literature.
(Hoerberger Piezo 2000) with an electrical energy Caldwell D.G. (1990a), "Compliant Polymeric Actuators
requirement of only 0.00 1W per valve, and a response as Robotic Drive Units", Ph.d Thesis, Dept. of
rate of 4O-50Hz which makes pulse modulated control Electronic Eng., University of Hull, Hull,
possible. E. Yorkshire, U.K.
Caldwell D.G. and Taylor P.M. (1990b), "Chemically
6.1 Actuator Test Results Stimulated Pseudo-Muscular Actuation", Int. J.
Of Eng. Science, pp.797-808, vol. 28, no.8.
The position regulation results for the control of the Caldwell D.G. (1992), "Polymeric Gels: Pseudo
proximal joint are shown in fig. 10, operating at an Muscular Actuators and Variable Compliance
input pressure of 250kPa. Tendons", IEEE/RSJ IROS '92 Conf, pp.950-57,

.:~~~~. ,~~ Raleigh, USA .

Caldwell D.G. and Goodwin M.(1993), "Pneumatic
Muscle Actuators for Control of a Dextrous
Manipulator", Intl. Conf. Advanced Mechatronics,
Tokyo, Japan.(accepted).

1 Vl lid :
~ 1 IU I U :
<u :1 I I)
Dario P., Beramasco M., Bemardi L., and Bicchi A.
(1987), "Shape Memory Alloy Actuating Module
for Fine Manipulation", IEEE Micro-robots and
Teleoperators Workshop, Hyannis, Mass.
l
~ ro i Nov.9-11.
~
O~~--~~--~--~-J
I Dynalloy (1989), "Biometal Guide Book", Dynalloy
Inc, Irvine, Cal., USA.
o 2 e e 10 12 Time (soc) Henke R.W. (1969), "Introduction to fluid Power
Fig.lO. System Response for the Proximal Joint. Circuits and Systems", Addision-Wesley.
Hu F.Q., Watson J., Payne P., and Page M. (1991),
These results show the potential of the controller "A fast Opto-pneumatic Converter for Robot
although enhanced control techniques will be needed in Actuation", lEE Colloquium on Robot Actuators.
producing truly efficient muscle actuator utilisation. Digest 1991/146, pp.10.1-2.
McKerrow P J. (1991), "Introduction to Robotics",
Addison-Wesley.
7. CONCLUSIONS Katchalsky A. (1949), "Rapid swelling and Deswelling
of Reversible Gels of Polymeric Acids by
For future generations of robots, especially those Ionization", Experimentia, Vol.5, pp.319-320.
required to work autonomously or semi-autonomously Scott D. (1985), "Electro-Rheological (ER) Fluid near
in non-factory environments, there are many key Commercial Stage", Auto. Eng .,Vol. 93, pp75-9.
sectors that require development, but among the most Takahashi S. (1987), "Piezo-electric Actuators and their
vital is the actuation system. Applications", J. Inst. Elect. In/. Comm. Eng.,
Vol 70, no.3, pp.295-7.
At present the most flexible model of a general Treloar L.R.G. (1975), "The Physics of Rubber
purpose actuator is provide by human muscle which Elasticity", Clarendon Press, Oxford.
has good power/weight output combined with Wassermann A. (1960),"Size and Shape Changes of
controlled action and robustness. Contractile Polymers", Pergamon Press, London.
Wolff H.S. (1970), "Biomedical Engineering", World
The braided muscle outlined in this report has very University Ubrary, London.
good power/weight (in excess of lkW/kg at 200kPa) Wilkie D.R. (1976), "Muscle", Edward Arnald, London.

527