A learning model for a neural oscillator to generate a locomotor pattern

Many living bodies have the ability to learn temporal signals, such as motor commands. Elucidation of the learning mechanism of such temporal signals is important for understanding information processing in the brain. It is known that dynamic components generate the movement of the many organs in living bodies. For instance, motor commands for many basic locomotor patterns such as walking and swimming are generated by the central pattern generator (CPG) which is composed of collective neural oscillators [1{3]. Heartbeat and peristaltic movements are also generated by coupled oscillatory components. The author has investigated learning models of temporal signals in coupled phase oscillators [4{7]. In this article, the proposed learning rule for acquisition of an adequate phase relation between two phase oscillators according to an evaluation function [5] is brie y summarized in section II. In section III, a learning rule for a neural oscillator is derived. Simulation results, including the adaptive control of a zerolegged robot, are shown in section IV.