An internal sensory model allows for balance control based on non-actionable proprioceptive feedback

1 Mar 2024  ·  Eric Maris ·

All motor tasks with a mechanical system (a human body, a rider on a bicycle) that is approximately linear in the part of the state space where it stays most of the time (e.g., upright balance control) have the following property: actionable sensory feedback allows for optimal control actions that are a simple linear combination of the sensory feedback. When only non-actionable sensory feedback is available, optimal control for these approximately linear mechanical systems is based on an internal dynamical system that estimates the states, and that can be implemented as a recurrent neural network (RNN). It uses a sensory model to update the state estimates with the non-actionable sensory feedback, and the weights of this RNN are fully specified by results from optimal feedback control. This is highly relevant for muscle spindle afferent firing rates which, under perfectly coordinated fusimotor and skeletomotor control, scale with the exafferent joint acceleration component. The resulting control mechanism balances a standing body and a rider-bicycle combination using realistic parameter values and with forcing torques that are feasible for humans.

PDF Abstract
No code implementations yet. Submit your code now

Tasks


Datasets


  Add Datasets introduced or used in this paper

Results from the Paper


  Submit results from this paper to get state-of-the-art GitHub badges and help the community compare results to other papers.

Methods


No methods listed for this paper. Add relevant methods here