Modular control of limit cycle locomotion over unperceived rough terrain

We present a general approach to design modular controllers for limit cycle locomotion over unperceived rough terrain. The control strategy uses a Central Pattern Generator (CPG) model implemented as coupled nonlinear oscillators as basis. Stumbling correction and leg extension reflexes are implemented as feedbacks for fast corrections, and model-based posture control mechanisms define feedbacks for continuous corrections. The control strategy is validated on a detailed physics-based simulated model of a compliant quadruped robot, the Oncilla robot. We demonstrate dynamic locomotion with a speed of more than 1.5 BodyLength/s over unperceived uneven terrains, steps, and slopes.