Movement execution and control (wk1) Flashcards
What are the 3 brain areas mainly involved in movement execution and control?
-Primary motor cortex (controls vision)
-PRR, area LIP, area AIP
-Mirco-stimulation of neurons in monkey’s primary motor cortex can elicit voluntary motor behaviours
What is the main observation of Graziano 2005’s experiment?
What is the main observation of Graziano 2005’s experiment:
-Execution and control -> Purposeful reaching movements are generated by stimulating reach-related neurons. Neurons tend to code the final reach position, rather than the starting position (Graziano et al, J Neurophysiol 2005)
What is the directional tuning of motor neurones (Georgeopolous et al. 1986)
-Neurons in primary motor cortex have their own preferred directions (directional tuning).
What kinds of human movement are controlled by feedforward control?
-Movements which are planned ahead (pre-decided)
-Ballistic, open-looped control is needed to generate very quick movement, when there is no time to process the sensory feedbacks.
-Typical sensorimotor delay is ~200 msec. Fast eye movements. Called saccades last only <200 msec.
-There is no time to accommodate feedback information, which means saccades are controlled feedforwardly.
Describe feedforward control in motor control and latency
-Feedforward control is also needed for movement that required very little latency: Vistibulo-Ocular Reflex (VOR) -> Typical latency of VOR is 8~9 msec, initiated way earlier than visual feedback kicks in. VOR is not programmable, unlike saccades.
-Feedforward in motor control -> It is mostly used for eye movement whose inverse model is relatively simple. Feedforward control is used in the initial part of the movement, before the sensory feedback kicks in. For movements with more complicated dynamics, or to control the later part of the movement, we do need feedback controllers.
What are the problems of sensorimotor delays?
- Delay in visual feedback -> ~100ms to be processed in the retina and transmitted to visual cortex and Central processing (among neurons) add another ~100ms
- Delays in muscle feedbacks (reflexes) -> 10-40ms before a muscle spindle signal reaches CNS
-Delayed feedback is often unstable and uncontrollable
What are the strategies to compensate for sensorimotor delays?
- Intermittency -> Pause until sensory feedback arrives and then resume – saccades, manual tracking and balancing
- Prediction compensates for sensorimotor delays -> If you know what is going to happen in the future, as a consequence of your own action, you can proactively control your movement based on the prediction
How is the forward model used for anticipatory feedback control?
-Forward model enables anticipatory control -> The forward model predicts the consequence of action based on the efference copy. This prediction can be done with a significantly less delay – you can anticipate what sensory feedback will be received before it arrives.
-Early errors made during reaching are corrected quickly before the sensory feedback arrives (~45 ms after the movement onset)
In which situations are the feedforward and feedback controllers used together?
-Due to the trade-off between feedforward and feedback, the brain incorporates both control strategies (in sequence)
-Predictive (anticipatory) control needed for quick reaction
-Feedback adjustment needed for flexible tuning
What can sensory feedback from the observer model be split into 2 with?
- Sensory feedback that I already expected: no surprise
- Sensory feedback that I didn’t expect: surprise (also called innovation)
How can the mismatch between prediction and sensory feedback arise?
-with the observer model
- The forward model is wrong
- The sensory feedback is corrupted by noise
What is the observer model and the forward dynamic model?
Observer model -> How to mix what we predicted and what was sensed
-Forward dynamic model -> What is my current status (position velocity, joint, angles etc)
What is the forward sensory model?
-Forward sensory model -> In that status what sensory information (tactile, sensation, gravity etc) I am supposed to sense?
-The observer model optimally mixes predicted and actual sensory feedback based on their uncertainty