Implementation for autonomous cooperative motion by the recurrent neural networks using moderatism

摘要

A novel learning rule is proposed to adopt an appropriate evaluation of the Recurrent Neural Network system controlled by Moderatism. This learning rule uses dynamic correlation between two contiguous neurons. The system composed of some neuron oscillators behave appropriately owing to cooperation of neuron oscillators. This learning rule is verified through computer simulation and experiments using a snaking robot with multiple joints.