Lecture 19 - computational principles of motor control

Question 1

What is David Marr’s 3 levels of analysis?

Answer 1

• implementational (what is the system made from?)
• algorithmic (what does the system do?)
• computational (what problem does it solve?)

Question 2

How you would describe the control of multi-joint movements?

Answer 2

• straight trajectories (straight line from start to end point)
• bell shaped velocity profiles (start with slow speed and peaks in the middle)

Question 3

Feedback control with time delays - describe low gain.

Answer 3

• takes a long time to get to the target
• think taps in a house you don’t know, slowly increase the temperature until you reach the goal

Question 4

Feedback control with time delays - describe high gain.

Answer 4

• doesn’t take time to reach the goal but actually exceeds the goal so then you try to counteract this and go below the goal again, shows oscillations
• think taps again, very quickly increase the temperature and it gets too hot, so you decrease the temperature and it goes too cold again.

Question 5

Describe internal models for motor control.

Answer 5

• motor system uses knowledge about the body and environment to make fast, accurate, feedforward movements even when feedback is slow
• these need to be adaptable in the face of a changing environment

Question 6

What is the bean bag experiment used for?

Answer 6

• used to explain adaptation internal models
• brain is adapting to the visual change it sees
• when the visual change is corrected the brain still tries to throw the bean bag inaccurately as it is still thinking it needs to be adapted.

Question 7

What drives adaptation?

Answer 7

discrepancy between expected and actual consequences

Question 8

What is the feedforward model used for?

Answer 8

to predict the consequences of actions.
used to distinguish between self-generated and external sensory feedback

Question 9

What do you use an inverse model for?

Answer 9

assess the consequences of the movement to make corrections

Question 10

What is optimal state estimation?

Answer 10

• combines internal sensory predictions with external sensory feedback to generate beliefs about the state of the body and the world

Question 11

What can improve the forward model?

Answer 11

discrepancies between the internal predictions and external feedback

Question 12

What are the 4 different redundancy and cost functions?

Answer 12

• 2 agonist muscles used together to generate a particular force
• many different strategies (one muscle does all the work, the other doesn’t and vice versa)
• the inaccuracy of both muscles will be combined with variances added
• inaccuracies are minimised by distributing effort between both muscles

Question 13

What is the uncontrolled manifold hypothesis?

Answer 13

• slightly different movements can achieve the same goal (redundancy).
• together these movements define the uncontrolled manifold.
• skilled movement exploits variability in the uncontrolled manifold to improve accuracy in task-relevant dimensions

Question 14

What is the minimum intervention principle?

Answer 14

• deviations are only corrected when they interfere with the task goals
• minimum intervention reduces signal-dependent errors caused by the correction itself
• for complex movements, task-relevant and task-irrelevant dimensions may be complicated to work out