A Novel Dynamic Locomotion Control Method for Quadruped Robots Running on Rough Terrains

Quadruped robots have excellent application prospects whereas the locomotion control of them on rough terrains is still a challenging problem, especially for those of large scales. The existing methods are either too complicated or lack of accuracies due to assumptions used. This paper presents a novel control algorithm for quadruped robots running on rough terrains inspired by the virtual model control and the model predictive control. State recursions are carried out based on the dynamic model of the trunk during the standing phase. The modeling of the body is implemented in the self-defined motion reference frame that avoids global state estimations and accumulative errors. The force distribution of the standing legs is realized by quadratic optimization involving state predictions. Forces of the swing legs are calculated by the virtual spring-damping model that follow the desired trajectory which is robust to external disturbances. These two sub-controllers are combined by the time-force based state machine. Simulation results show that the quadruped robot obtains the adaptability to rough terrains and robustness to lateral pushes with the proposed method.