MURI

1
Biomimetic Robots
MURI
2
Comparison with Artificial Muscles
New Results on Measurements of Muscles
Fabrication
Gecko foot adhesion
Discussion of low level mechanism
3
MURI Year Two Meeting2000
Professor Robert J. Full Daniel Dudek Dr. Kenneth
Meijer
Basic properties of natural muscle
First direct comparison of natural muscle to
artificial muscle
Fabrication
Diverse roles of muscles
4
Manufactured Legs
SDM permits embedded sensors and actuators
What properties should legs possess? Why? What
properties should the actuators possess? How many
actuators should there be? How should the
actuators be controlled?
5
MURI Interactions
Rapid Prototyping
Stanford
Muscles and
Motor Control
Learning
Locomotion UC Berkeley
Johns Hopkins
MURI
Robot Leg Mechanisms
Manipulation
Harvard
UC Berkeley
Sensors / MEMS
Stanford
6
Interdisciplinary Collaboration
Actin/Myosin
Ion Channels
CPG
Metabolic Pathways
Ion Channels
Proteomics
General Biological Principles
Novel Hypotheses Devices
Biological Inspiration
General Robot Design Principles
7
Road Map
1. What muscles can do. (Traditional
characterization) 2. What muscles do in nature.
(Inputs values from behavior)
3. Compare natural
muscles to artificial muscles.
8
Road Map
1. What muscles can do. (Traditional
characterization) 2. What muscles do in nature.
(Inputs values from behavior)
3. Compare natural
muscles to artificial muscles.
9
Muscle Model
10
Activation
Human
Muscle Force
Stimulation (EMG)
Cockroach
11
Activation
Time to Peak Force 0.004 - 0.79 sec 200-fold
variation Time to 50 Relaxation 0.009 - 1.1
sec 100-fold variation
600
Maximum isometric stress 7 - 803 kN/m2 or
kPa 100-fold variation
Insect leg muscle
500
400
Force (mN)
300
200
100
0
0
20
40
60
80
100
Time (msec)
12
Muscle Model
13
Force-Length Curve
Maximum isometric stress varies with Strain
Animals tend to operate on the Ascending or
Plateau region.
14
Force-Length Variation
Maximum Strain varies from 2 - 200 100-fold
variation
Relative Stress ()
Strain
15
Muscle Model
16
Force-Velocity Curve
Maximum Contraction Velocity 0.3 - 20
l/sec 60-fold variation
17
Force-Velocity Curve
Trade-off between Force and Velocity Similar
Shape of Curve
18
Instantaneous Muscle Power
Maximum Instantaneous Power Output at 1/3 Maximum
Contraction Velocity
Power Force X Velocity
Muscle Force
Muscle Velocity
19
Instantaneous Muscle Power
Maximum Instantaneous Power Output gt 500 W/kg
muscle
20
Road Map
1. What muscles can do. (Traditional
characterization) 2. What muscles do in nature.
(Inputs values from behavior)
3. Compare natural
muscles to artificial muscles.
21
In Vivo Activation
Muscles Activated Rhythmically at a Given Phase
22
Cycle Frequency
1000
Mosquitoes
F
l
i
e
r
s
Flies
Flower flies
Bees, Wasps
Aphids,
100
Crane flies
White flies
Beetles
H
z
Dragonflies
Frequency lt1 to 1000 Hz
Sphinx moths
Butterflies
Saturnid moths
10
R
u
n
n
e
r
s
S
w
i
m
m
e
r
s
Invertebrates
Full, 1997 Handbook of Comparative Physiology
1
m
0.1 mg
1 mg
10 mg
0.1 g
1 g
10 g
0.1 kg
1 kg
10
g
Body mass
23
Muscle Lever
Control
Stimulation
Stimulation
Servo and
Strain
Force
Transducer
Frequency
24
Workloop Technique
Lever
25
Workloop Technique
26
Muscles as Motors
Power Generation 9-284 W/kg
27
Workloop Shape
Shape depends on Frequency
28
Power vs Frequency
Work per cycle decreases with Frequency Power
constant
Scallop
Bee
29
Stress, Strain vs Frequency
Stress and Strain decrease with Frequency
30
Road Map
1. What muscles can do. (Traditional
characterization) 2. What muscles do in nature.
(Inputs values from behavior)
3. Compare natural
muscles to artificial muscles.
31
Artificial Muscle?
First Direct Comparison by K. Meijer
Artificial Butterfly
Collaboration
S. V. Shastri R. Kornbluh R. Pelrine
Acrylic Dielectric Elastomer
SRI research engineer Roy Kornbluh
32
Dielectric Elastomer Actuators

• Soft ElectroActive Polymers (EAP)
• Polymer film is sandwiched between compliant
  electrodes and acts as a dielectric (insulator).
• Incompressible polymer gets thicker and contracts
  in area when a voltage is turned off.

Basic functional element

33
Activation
EAP has Rapid Kinetics
34
Force-Length Curve
EAP has a linear Force-Length Curve
35
Acrylic Dielectric Elastomer
Same Apparatus used to test Natural Muscle
Force
Dlength
46.2 mg at a 1 N pre-tension Dimensions of active
part of the actuator (l x w x h) 17.88 x 15.88 x
0.07 mm.
36
Power Output
EAP Produced and Absorbed Energy
37
EAP Power Output
As in Muscle, EAPs only Produce Power over a
Particular Range of Strains and Stimulation Phases
38
Work vs Frequency
Work per Cycle Lower than mean Activation not
Maximal
39
Stress, Strain vs Frequency
Stress higher and Strain lower than mean.
40
Power Output Comparison
EAP
EAP within Range of Natural Muscle
41
Conclusions
1. Muscles have a broad range of potential
function. 2. Matching natural inputs required to
reveal function 3. Can not refute EAP as
artifical muscle
42
Actuator Performance Comparison
43
Actuator Performance Comparison - Stress vs.
Strain
44
MURI Year Two Meeting2000
Professor Robert J. Full Dr. Anna Ahn Dr. Kenneth
Meijer
Basic properties of natural muscle
First direct comparison of natural muscle to
artificial muscle
Fabrication
Diverse roles of muscles
45
Multiple Muscle Systems
Complex, Redundant? or Diverse Functional
Capacity?
46
Questions
Why are there so many muscles operating at a
single joint? Are all muscles created equal? Can
differences in function be explained by neural
activation alone? Can differences in function be
explained by traditional characterizations? Are
muscles mainly power generators?
47
Hypotheses
Muscles of the same anatomical group activated at
the same time will function similarly.
48
Two extensor muscles innervated by a single motor
neuron
muscle 178
coxa-femur joint
muscle 179
stance phase
joint extension muscle shortening
small joint angle long muscle lengths
large joint angle short muscle lengths
Anna Ahn
49
Hypothesis Muscles stimulated by the same motor
neuron function similarly.

• NEURAL CONTROL
• Stimulation patterns the same?
• INTRINSIC MUSCLE PROPERTIES
• Force-Length properties similar?
• Force-Velocity properties similar?
• Twitch kinetics similar?
• Shortening deactivation similar?

50
Stimulate motor neuron, while measuring EMGs
from 178 and 179.
(mean S.D.)
51
Hypothesis Muscles stimulated by the same motor
neuron function similarly.

• NEURAL CONTROL
• Stimulation patterns the same? YES
• INTRINSIC MUSCLE PROPERTIES
• Force-Length properties similar?
• Force-Velocity properties similar?
• Twitch kinetics similar?
• Shortening deactivation similar?

52
Similar force-length properties
179
178
53
Hypothesis Muscles stimulated by the same motor
neuron function similarly.

• NEURAL CONTROL
• Stimulation patterns the same? YES
• INTRINSIC MUSCLE PROPERTIES
• Force-Length properties similar? YES
• Force-Velocity properties similar?
• Twitch kinetics similar?
• Shortening deactivation similar?

54
Similar force-velocity properties
55
Hypothesis Muscles stimulated by the same motor
neuron function similarly.

• NEURAL CONTROL
• Stimulation patterns the same? YES
• INTRINSIC MUSCLE PROPERTIES
• Force-Length properties similar? YES
• Force-Velocity properties similar? YES
• Twitch kinetics similar?
• Shortening deactivation similar?

56
Similar isometric contraction kinetics
178
179
57
Hypothesis Muscles stimulated by the same motor
neuron function similarly.

• NEURAL CONTROL
• Stimulation patterns the same? YES
• INTRINSIC MUSCLE PROPERTIES
• Force-Length properties similar? YES
• Force-Velocity properties similar? YES
• Twitch kinetics similar? YES
• Shortening deactivation similar?

58
Similar shortening deactivation
178
179
(mean S.D.)
59
Hypothesis Muscles stimulated by the same motor
neuron function similarly.

• NEURAL CONTROL
• Stimulation patterns the same?
• INTRINSIC MUSCLE PROPERTIES
• Force-Length properties similar?
• Force-Velocity properties similar?
• Twitch kinetics similar?
• Shortening deactivation similar? YES

60
Muscle Power during Running
Two extensor muscles at same joint stimulated by
the SAME neuron have different function.
Stiffening Element
Damper or brake
3 W kg-2
-19 W kg-2
stimulation
61
Whats different?
62
Active force during shortening
178
179
stance
stance
stimulation
63
Conclusions

• 1. Muscle function cannot be predicted from
  neural activity.
• Muscles innervated by the same motor neuron do
  NOT necessarily function similarly.
• 2. Muscles of the same anatomical group (178 and
  179) can have many similar intrinsic muscle
  properties, but still function differently.
• 3. History-dependent properties may play an
  important role in determining muscle function.

64
Implications for Robotics

• 1. Direct copying of the musculoskeletal system
  is likely to fail.
• Muscle have diverse roles that can only be
  revealed by extensive experimentation.
• 2. Control and energy management may be attained
  using actuators with different properties rather
  than sending out complex control signals.
• 3. EAPs with muscle-like properties are
  available. More direct comparison are needed.
  More emphasis on function in devices required.

65
Robotic Applications of EAPs
Modular design composed of individual stretched
film actuators integrated into a 6-legged walking
robot
CAD representation of the robot including a
second degree of freedom