Neural network models of the brain mechanisms of bilateral coordination
David Scott Farrar, David Zipser · 1999
This thesis couples analytical techniques from theoretical robotics with neural network modeling to investigate possible brain mechanisms underlying the coordination of bilateral movements. It presents several neural network models that are biologically plausible candidate mechanisms for the neural computations underlying coordination. The models provide a connectionist account of how the brain might accomplish coordinated bilateral movements. The thesis begins by describing several mechanisms for controlling the movement of a pair of arms. The first is an engineered motion planner that finds solutions to the ill-posed problem of making non-colliding, goal-directed movements. The planner exploits a representational tool that does not require a forced decomposition of coordination problems. The second uses feedforward neural networks that learn to emulate the coordinated behaviors of the motion planner using considerably less computational resources. Knowledge about the structure of the body necessary to solve the coordination problems is learned by the neural networks, and comes to be represented in the internal weights. The third uses biologically inspired recurrent network architectures to more closely approximate the kinds of neural phenomena seen in the brain. Analysis of the networks shows in general terms how they work, and allows us to make testable predictions about some of the response properties that might be observed in the brain systems serving bilateral coordination.