Transcript Motor Synergies: A Concept in Motor Control

Motor Synergies:
A Concept in Motor Control
Marieke Rohde
CCNR – Centre for Cognitive Neuroscience and
Robotics
Workshop on the Dynamical Systems
approach to Life and Cognition
University of Sussex, 8 and 9 March 2005
Structure
1. Motor Synergies
2. Directional Pointing
•
•
Linear Synergies in Human Directional
Pointing
Evolutionary Robotics Example
3. Conclusions
1.) Motor Synergies
Nicholas Bernstein
• Nicolas Bernstein (1967,
but really 1935)
– Physiology of Activity,
Biomechanics
– Degrees of Freedom
Problem
The Degrees of Freedom Problem
• The “Cartesian
Puppeteer” has to
control a countless
number of motor
units.
DoF
2600
26
7
Motor Equivalence and ContextConditioned Variability
• Motor Equivalence
– Redundancy through many
degrees of freedom
• Context-Conditioned
Variability:
– Anatomical (role of a muscle is
context dependent)
– Mechanical (command sent to
muscles is ignorant against
motion/nonmuscular forces)
– Physiological (the spinal cord is
not just a relay station)
The Solution
• Systematic relationships
between actuators (constraints)
can reduce the degrees of
freedom to form functional motor
units (e.g. wheel position in a
car)
 Motor Synergies!
• Skill Acquisition
– First freezing degrees of freedom
– Then freeing them and exploiting
passive dynamics
Biological Evidence for Synergies
• Systematicities in
kinetics/kinematics:
– Different types of gaits
– Shooting
– Breathing
(Overview: Tuller et. Al. 1982)
– Linear relation between
shoulder and elbow torque
(Gottlieb et. Al. 1999)
• Complex behaviour as
composition of synergies
– Frog EMG data can be
explained as linear
combination of 7 linear
synergies (Saltiel et. Al. 2001)
Synergy between elbow
and shoulder joint in a
skilled marksperson
Problems with Motor Control
through Synergy Control
• The reminder of the homunculus
– How does it work?
• Acquisition and maintenance of synergies:
– What is a good synergy?
– What mechanism controls their development?
• Combination of synergies:
– Who deals with non-linearities?
• Weiss, P. and M. Jeannerod (1998):
– “Motor coordination is not the goal but a means to
achieve the goal of an action”
If there’s no homunculus…
…there’s no problem.
Still, the observed phenomena require explanation.
Evolutionary Robotics
• Things we can ask ourselves:
– What does it imply if we have non-redundant models?
– What does it imply if we do not have context-conditioned
variability?
• Things we can investigate:
–
–
–
–
More degrees of freedom
Anatomical, mechanical, physiological context dependence
Motor synergies in the absence of a homunculus
Impact of these factors on
• Behaviour
• Evolvability
• The “phylogenetic learning” process
2.) Directional Pointing
Linear Synergies in Human
Directional Pointing
• Gottlieb et. Al. 1997:
– Directional Pointing in the sagittal
plane
– Linear relation:
– Systematic variation of scaling
constant with pointing direction
– Linear synergies as an outcome of
learning?
Hand trajectories for pointing
Direction against scaling constant
• Zaal et. Al. 1999:
– Linear Synergies are not learned,
they constrain learning
Pre-reaching period
Experiments (Work in Progress)
• Simulated Robotic Arm
• Proprioceptive (joint angle) +
Directional Inputs
• Fitness: Position at endpoint
• Motor control:
– “Garden CTRNNs” with two motor
neurons per degree of freedom
– “Split Brain CTRNNs” with separate
controllers for joints
– Linear Synergy networks with just one Screenshot of the simulated arm
motor output and evolved scaling
function (RBFNs)
– 2 vs. 4 degrees of freedom for all of the above
– 2 goals vs. up to 6 goals (additional goal once it has a certain level)
• Most severe simplifications:
– Hand of 4 degrees model is squashed between two planes
– No gravity
Results
• Performance
– DoFs: 4 twice as good as 2
– Linear synergy are much better than CTRNNs (even linear linear
synergies are comparable)
– Split brains are not a lot worse than ordinary CTRNNs
• How do they solve the problems?
–
–
–
–
2 DoF’s: Frequently just use one joint
4 DoF’s: exploit the invisible planes.
Linear Synergy: use different techniques, look a bit smoother
Split brains: Independence of joints very obvious
Results
• Phylogeny
– CTRNNs freeze degrees of
freedom at first and then
include them.
– Networks use passive
dynamics straight away.
• Synergies
– No linear synergies in any
CTRNN controllers.
3.) Conclusions
Conclusions: Evolutionary Robotics
• Redundant degrees of freedom can facilitate evolving a
controller, in spite of the much bigger search space and
lead to a better solution
• Learning under the constraint of linear synergy reshapes
the search space and can lead to a very quick and
successful evolution of different strategies (Careful with
bias through model selection).
Conclusions: Synergies
• A question we cannot answer (yet) is: Why are there linear
synergies?
• The acquisition of synergies:
– Learning is not necessarily building up linear synergies.
– The fact that the constraint of linear synergy boosts evolution suggests
its suitability for developmental processes.
– CTRNNs freeze and free DoFs.
– Some kind of synergy gives CTRNNs an advantage over split brain
CTRNNs.
• The concept of synergy:
–
–
–
–
It is very useful to explain behaviour in abstract terms.
Particularly, if more complex behaviour is investigated.
Thinking in terms of synergies raises different questions
You just have to be clear about your relation to the Homunculus idea.
Future Research
• Input models:
– Make more CTRNN friendly
– Visual inputs
• Get rid of the invisible planes
• Evolve constraints for lifetime development
• Use synergies in a larger context (co-evolution of car
and driver)
• Investigate other forms of context conditioned variability
Any questions?
References
•
•
•
•
•
•
Arbib, M. A. (1981): Perceptual Structures and Distributed Motor
Control. In: V. B. Brooks (ed.): Handbook of Physiology. Section 2: The
Nervous System. Vol. II, Motor Control, Part 1. American Physiological
Society, 1449-1480.
Bernstein, N. (1967): The Coordination and Regulation of Movements.
Oxford: Pergamon.
Berthouze, L. and M. Lungarella (2004): Motor Skill Acquisition Under
Environmental Perturbations: On the Necessity of Alternate Freezing and
Freeing of Degrees of Freedom. Adaptive Behavior, 12(1).
Gottlieb, G. L., Q. Song, G. L. Almeida, D. Hong, and D. Corcos (1997):
Directional Control of Planar Human Arm Movement. Journal of
Neurophysiology 78:2985-2998.
Grossberg, S. and Paine, R.W.(2000): A Neural Model of Corticocerebellar
Interactions During Attentive Imitation and Predictive Learning of Sequential
Handwriting Movements. Neural Networks, 13, 999-1046.
Morasso, P., F.A. Mussa Ivaldi and C. Ruggiero (1983): How a
discontinuous mechanism can produce continuous patterns in trajectory
formation and handwriting. Acta Psychologica 54. pp. 83-98.
References
•
•
•
•
•
•
Sporns, O., and G.M. Edelman (1993): Solving Bernstein's problem: A
proposal for the development of coordinated movement by selection. Child
Dev. 64:960-981.
Saltiel, P., K. Wyler-Duda, A. d'Avella, M.C.Tresch and Bizzi, E. (2001):
Muscle Synergies Encoded Within the Spinal Cord: Evidence From Focal
Intraspinal NMDA Iontophoresis in the Frog. J. Neurophysiol., 85: 605-619.
Tuller, B., H. Fitch and M. Turvey (1982): The Bernstein Perspective: II. The
Concept of Muscle Linkage or Coordinative Structure. in: S. Kelso (ed.):
Human Motor Behavior. An Introduction. Hillsdale: Lawrence Erlbaum.
Turvey, M., H. Fitch and B. Tuller (1982): The Bernstein Perspective: I. The
Problems of Degrees of Freedom and Context-Conditioned Variability. in: S.
Kelso (ed.): Human Motor Behavior. An Introduction. Hillsdale: Lawrence
Erlbaum.
Weiss, P. and M. Jeannerod (1998): Getting a Grasp on Coordination. News
Physiol. Sci. 13. 70-75.
Zaal, F., Daigle, K., Gottlieb, G.L., Thelen, E. (1999): An unlearned principle
for controlling natural movements. Journal of Neurophysiology, 82:255-259.