Neuro-control using temporal dynamics for robot locomotion

摘要

A neural controller for an autonomous walking machine called theMicro-Rover is presented. Dynamic properties of real neurons wereincorporated within the neural model in the form of delayed excitation.The pattern generation capability of networks of biological neurons werestudied in order to choose an appropriate gait controller for theMicro-Rover. A four-phasic network with recurrent inhibition was foundto be suitable for generating the locomotion rhythm. Sensory modulationof the locomotion rhythm takes place within local networks thatcoordinate leg and frame actuation. A simple reflex circuit enables theMicro-Rover to avoid obstacles. The networks were developed and testednumerically, and work is currently underway to implement the controlleron the Micro-Rover