Neural Control of Reaching

Principal Investigator: Amy Bastian

Aim1: Use dynamic computational models of the human arm and a robot exoskeleton to investigate the mechanisms of cerebral versus cerebellar coordination deficits. 

Cerebral coordination deficits from hemiparesis are thought to arise from development of abnormal synergies (i.e. abnormal co-activation of muscle groups). We will determine whether there are dominant synergies in subjects with motor hemiparesis from cerebral damage and if they predict the pattern of reaching deficits observed. 

Cerebellar coordination deficits are hypothesized to be due to either 1) mis-estimation of limb dynamics, leading to poor interaction torque compensation (i.e. torques occurring at one joint due to motion of another) or 2) muscle timing delay. Here we will examine the degree to which the dysfunction of an internal dynamic model describes the deficits observed in people with cerebellar damage and compare it against the other leading theory, muscle activation delay. 

Aim 2: Use a robotic exoskeleton for the arm to modify the appropriate mechanical parameters identified in Aim 1 to achieve a more normal movement trajectory through either motor learning or compensation mechanisms. We will use a Kin Arm robot that allows us to directly influence individual arm joints to investigate the two general approaches. 

Approach 1: We hypothesize that short-term learning (i.e. adaptation) can be improved in cerebellar and cerebral subjects using a targeted control law that matches the deficit found in Aim 1. We will study whether a control law that exaggerates human control error is beneficial in teaching the subject how to counteract that error, in comparison to a conventional control law that guides the subject through the correct motions. 

Approach 2: For subjects that do not learn using Approach 1, we hypothesize that a targeted control strategy that compensates for human control errors will yield smooth and coordinated motion. This strategy effectively “adds back in” the aspects of limb control that have been lost for each subject. We will also test three more general, but potentially useful, compensation strategies: 1)a viscous damping field, 2)a force channel, and 3)a pseudo-admittance control law. These compensations will be compared to each other and to a no force (null) control condition.