A learning model for oscillatory networks
Neural Networks
Introduction to Reinforcement Learning
Introduction to Reinforcement Learning
Fast learning in networks of locally-tuned processing units
Neural Computation
Constraint and exploitation of redundant degrees of freedom during walking
Robotics and Autonomous Systems
| Hi-index | 0.00 |
Many cyclic movements of vertebrates are produced by a rhythm generator in the spinal cord, which is often referred to as the central pattern generator. Meanwhile, higher centers, such as the cerebellum, are also involved in motor control. In this study, we discuss the control and learning mechanisms of these two control systems, focusing on the following problems: (1) how these two control systems generate a motor command cooperatively without conflict, and (2) how these control systems acquire motor commands. As a possible model to solve these problems, we propose a hierarchical motor control learning model on the basis of physiological knowledge. This model assumes that the higher center learns the timing and amplitude of a motor command and sends control signals to the rhythm generator. The rhythm generator is tuned by the control signals and acquires the ability to output a motor command to realize a target motion. By applying this model to a simulation experiment of the learning control of a one-dimensional hopping robot, hopping at a desired height was successful. Our simulation results also show that a multiple control system involving the higher center and rhythm generator is more robust than a single control system.