W7&8 Control Strategies

Question 1

What is meant by a strategy?

Answer 1

an muscle activation pattern that minimises/maximises a task relevant cost function

Question 2

what is meant by a cost function?

Answer 2

• a function that determines how effective/ineffective a movement is
• e.g. 100m sprint
  -> cost function is running speed speed then you’re trying to maximise that
  -> cost function is time to complete race then you’re trying to minimise that

Question 3

examples of strategies

Answer 3

• minimise effort
• minimise muscle activations
• minimise jerk - jerk is the rate of change of acceleration
• maximise performance outcome
• maximise likelihood of ‘success’

Question 4

define skill

Answer 4

“the ability to bring about some end result with maximum certainty and minimum outlay of energy” (Schmidt)

Question 5

typical research methods of determining control strategies people are using

Answer 5

1. Record performance (experimental biomechanics)
2. Simulate performance (theoretical biomechanics)
3. Optimise technique for various criteria
4. Select solution that best matches recorded one

Question 6

simulation models

Answer 6

• based on Newton’s equations of motion
• inputs –> model –> outputs

Question 7

draw dynamical systems theory (Newell, 1986)

Answer 7

Question 8

example constraints in a modelling context using dynamical systems theory framework

Answer 8

organismic/individual:
- Stature
- Mass
- Segmental inertia properties
- Prevent hyperextension, etc.
environmental:
- gravity
- ground interaction
- interactions with equipment
task:
- bilateral symmetry
- velocity
- duration
- balance
- secondary objectives

Question 9

local or global optima

Answer 9

• The y axis might be energy cost or time taken to run a race (trying to minimise it)
• X axis would be our cost function e.g. knee angle
• As you change it the cost function variable, the result of the cost function also changes
• If it’s not a straight line you’ll get lots of local optima and then one global optima
• We want to find the global optima
• But we might get stuck in a local optima
• Human body is good at making small changes and finding local optima but not as good at finding global optima
• Each local optima might be the best way of achieving a different constraint
• Wider trough = more room for error

Question 10

Walking example of control strategies

Answer 10

(Gaesser et al., 2022)
- Participants walking in 3 different ways and looking at energy expenditure
- Humans usual walk has a relatively low energy expenditure
- Putey walk didn’t make much difference
- Teabag walk made a massive difference and resulted in much bigger energy cost

Question 11

Walking gait simulation model

Answer 11

(Ren et al., 2007)
- Relatively simple model of someone walking
- 4 different solutions - each of these would be a local optima
-> Stiff-knee gait
-> Inadequate knee extension
-> Excessive ankle plantar flexion
-> Predicted gait with lowest energy consumption - this one looks like humans normal gait suggesting humans have learned to in a way that minimises our energy cost

conclusions:
- Multiple solutions for the given constraints
- Minimum energy closely resembled human gait
- Possible other constraints not considered
- When there is a choice of multiple solutions, skilled performance often uses the one that results in the minimum outlay of energy

Question 12

Gait transitions - horse example

Answer 12

(Hoyt & Taylor, 1981)
- Looking at horses moving in 3 ways: walk, trot, gallop
- For each movement type there’s an optimum speed where it’s most efficient
- E.g. if a horse tries to trot really slowly or quickly it’s not efficient
- This study suggests energy cost serves as a trigger to change gait
- This corresponds well with self-selected speed in humans

Question 13

cross country skiing example

Answer 13

(Herzog et al., 2015)
- Two different ways of skiing - one arm swing per leg or double arm swing per leg
- At low speeds a 1-skate is more efficient (lower energy cost)
- At faster speeds a 2-skate is more efficient

Question 14

Gymnastics parallel bar undersomersault

Answer 14

(Davis, 2005)
- used to teach clear circle technique with little change in body shape
- but elite gymnasts started using stoop stalder technique with more hip flexion

Question 15

Gymnastics parallel bar under somersault simulation model

Answer 15

(Hiley & Yeadon, 2012)
- input: joint angle
- output: linear and angular momentum
- simulate to minimise joint torque and looked similar to old clear circle technique - so this technique has lowest energy cost
- but elite gymnasts are using stoop stalder technique - why?
- simulated to minimise horizontal velocity at release and looked similar to stoop stalder technique
- why do gymnast want to minimise horizontal velocity at release?
-> allows further skill development - almost stationary starting point for next skill
-> if they were to let go of the bar, gives them more margin for error
- this study shows that technique isn’t always based on minimising effort, especially when a performance outcome (task constraint) is more important

Question 16

Gymnastics upstart simulation - previous research - via-point

Answer 16

(Yamasaki et al., 2004)
- tried minimising jerk, minimising effort and minimising torque change
- regardless of what they tried they failed to match actual gymnasts movement without a ‘via-point’

Question 17

Gymnastics upstart simulation model

Answer 17

(Hiley & Yeadon, 2012)
- tried each combination 1,000 times adding noise to each joint angle
- noise scaled to variability in actual gymnast movement
- recorded number of successful simulations
- Minimising joint torque didn’t produce what gymnasts do
- Minimising torque change gets a bit closer but wasn’t exactly the same
- Maximising success made it look the same
- calculated joint angles RMS
-> squaring all errors, finding mean, square rooting mean

Conclusions
- Upstart technique characterised by maximising success (task constraints)
- Gymnasts have developed techniques that can cope with the level of noise within their movements

Question 18

Summary of W7 control strategies 1 lecture

Answer 18

• Choose movement patterns to maximise success
• When multiple techniques are viable, minimum energy cost solution often adopted
• Effort can be used as a trigger for technique change
• Minimising jerk or torque change can replicate muscle characteristics

Question 19

How does Bayesian Decision Theory link to movement?

Answer 19

• BDT is a framework aiming to select optimal actions based on inferences
• Wolpert (2007), describes how Bayesian statistics and decision theory is used in motor control to make movements with the most chance of success.

Question 20

Give an example of a movement which uses Bayesian Decision Theory

Answer 20

(Wolpert, 2007)
- Tennis plays often had prior knowledge of where to return the ball to.
- This just meant it was based on their own understanding.
- However, they often returned the ball to an area which was between their “prior” beliefs and the data which showed the best response.

Question 21

Bayesian approach

Answer 21

3 stages:
- Prior distribution is what you think/know prior to collecting data
- Data (sometimes called likelihood): update knowledge as more data becomes available,
- Posterior: final set of data with reduced uncertainty in our knowledge

Question 22

Why is it important to consider noise when selecting the most successful movement?

Answer 22

• Noise will always interfere with a movement and deter it from it’s desired outcome.
• Hiley & Yeadon (2012) study on gymnastics found that gymnasts performed best when considering noise and making small changes to their movements to become more successful

Question 23

What are some misinterpretations of p-values?

Answer 23

P-values give the probability of the data, not the null hypothesis. It is the probability of the observed outcome, and all more extreme outcomes, if the null hypothesis is true.

Question 24

Explain how the concept of signal dependent noise helps explain motor planning.

Answer 24

• Signal dependent noise is noise within a system that is dependent on the amplitude and frequency of the neural signal.
• Therefore a large force will create a large amount of noise.
• This may be part of the reason why humans usually seek out movements that expend the least possible energy.
• A movement using less energy and producing a smaller force will create less noise, therefore less error.

Question 25

Consequence of movements

Answer 25

(Wolpert, 2007)
- trying to catch a ball
- If you come down on the ball you’re not going to be able to catch it
- If you move upwards you can get there slightly late or slightly early you’ll be able to catch it still