24 - Models of Motor Control

Summary

This chapter focuses on the computational models of motor control. The current views of motor control are characterized by an awareness of the geometrical complexity of natural behavior and of the coordinate transformations involved in the production of movements. The discussion begins with a review of the geometrical framework of coordinate transformations involved in the mapping of motor plans into control signals. The key concept is that the mechanical properties of the musculoskeletal system provide a basis for simplifying the computational demands of this transformation. Bayesian statistics provide tools for describing the adaptive development of internal representations that map a desired movement into a motor command and, conversely, for estimating the state of motion of the body given a known motor command. The presentation ends with a discussion of neural network architectures capable of implementing various aspects of the computational tasks associated with the use of motor primitives.