A dynamic locomotion strategy for stair walking of a quadruped robot

In unstructured environments, blind walking can easily make the robot lose its balance or collapse due to unstable contact with the ground. One way to solve this problem is to adjust the gait period and the gait pattern. The gait period adjusts the speed of gait so it mitigates the amount of impact when the foot contacts the ground, and the gait pattern plays a role in stabilizing the balance of the base of the robot from external disturbances or impacts from the ground. Therefore, this paper suggests the gait controller using convex Model Predictive Control(MPC) with adjustable gait period and gait pattern. Since the controller uses the single rigid body model, it has a solid advantage that the computation time is relatively short than using the non-linear controller system. The stability of the controller has been verified by experiments that the robot without a vision system climbs and turns on the stairs with obstacles.