inverse models
Current position and desired position –> Motor commands.
- Used to create motor plans
forward models
Current position and Motor commands –> predicted position. Used to evaluate motor plans and/or action.
feedforward control
Motor command sent to muscle
Faster, but less accurate
feedback control
Motor command sent to muscle
Actual state compared to desired state, adjustments made based on errors
Slower, but more accurate
Premotor cortex role
- Involved in selecting goals and planning actions at a conceptual level, occurs before voluntary movement
- Particularly when plans are driven by external stimuli
Supplementary motor cortex (SMA) role
- Involved in selecting goals and planning actions at a conceptual level
- Particularly when plans involve internally generated sequences of actions (e.g. tying shoelaces)
Motor cortex/superior colliculus
- Motor cortex represents directional movements of body parts, not specific muscle actions
- Signals from motor cortex travel directly to lower motor and circuit neurons in brainstem and spinal cord
Basal ganglia role
- Help to select, initiate and inhibit movements through cortico-basal ganilia-thalamocortical loops
- Critical to dopamine-based reinforment learning
- Participate in motor, cognitive, and emotional control
Action initiation
Direct pathway
- Motor cortex excites striatum
- Striatum inhibits GP i/SNr
- GP i/SNr disinhibits thalamus
- Thalamus excites cortex
Population vector
- Accurately represents actual movement direction
- Treats firing of each neuron as a vector, then adds together
Direct pathway
- Motor cortex
- Striatum
- Globus pallidus pars interna/Sustantia nigra pars reticulata
- Thalamus
- Motor cortex
Indirectly pathway
- Cortex
- Striatum
- Globus pallidus pars externa
- Subthalamic nucleus
- Globus pallidus pars interna/Sustantia nigra pars reticulata
- Thalamus
- Cortex
Action inhibition
Indirect pathway
- Motor cortex excites striatum
- Striatum inhibits GPe
- GPe disinhibits STN
- STN excites GP i/SNr
- GP i/SNr reinhibits thalamus
Reinforcement learning
- Unexpected rewards
generate dopamine signals
from the substantia nigra
pars compacta (SNc) - This excites the direct
pathway (via D 1 receptors)
and inhibits the indirect
pathway (via D 2 receptors) - This allows modification of
behavior based on reward
Cerebellum role
- Uses forward model to
predict results of motor
commands - Uses differences between
actual results and
predicted results for:
●Online error correction
● Motor learning - Feedback control
Feedback control
- Feedback takes time
- The faster you go, the less time you have for
feedback - Less feedback leads to greater error
- This implies a speed/accuracy tradeoff
Fitts’s Law
T = a + b log2(D/W +1)
- For a target at distance D:
- Faster: if you want to reach the target in less time, T, then the target must get wider, W
-More accurate: if you want a narrower target, W, then the time must get longer, T
Local circuits in the spine can
- Control complex movements
- Respond to environmental changes
- Do not require higher-level input
Motor action and muscles
- Lower motor neurons synapse directly on muscle fibers
- Neurotransmitter causes muscle fibers to contract
- Muscle spindles detect changes in muscle length and send them back to spinal cord via dorsal root ganglia
Types of electrical recordings
- Intracellular
● Voltage clamp/Current clamp
● Patch clamp - Extracellular
● Single-unit recording
● Multi-electrode recording
● Field potentials - In vitro vs in vivo
- Anaesthetized vs awake
Brain-machine interfaces
- 2 microelectrodes arrays with hair-thin electrodes detect neuron signals are implanted into the left motor cortex
- Neuron signals pass to connectors attached to the skull
- Amplifies signals are fed to a brain-machine interface that interprets them
