Realization of the biologically-inspired dynamic walking controller of a quadruped robot

This paper introduces a control system design for an autonomous quadruped robot walking on irregular terrain. Biological cybernetics approach is used and the control of our walking robot can be considered hierarchical and modular. The upper computer functioning as the brain make decisions on what behavior is to be selected and the overall elements of its execution, such as the speed and the direction in the case of walking. The actual pattern activation of the legs of the quadruped robot is made locally by groups of neural oscillators termed CPGs (central pattern generators) implemented in the lower DSP (digital signal processor) controller. The peripheral feedback provides information from proprioceptor and exteroceptor. The output of CPGs can be controlled by higher level computer, while simultaneously modulated or adjusted by interactions between the CPGs upper controller and the sensory information from the robot body or circumstances. Consequently, locomotion of the quadruped robot is robust and coordinated.