Trajectory Adaptation and Safety Control via Control Barrier Functions for Legged Robots

摘要

Optimizing footholds as well as their exact transition trajectories during swing phase is computationally expensive hence can not satisfy the real-time requirements when traversing challenging terrain. This paper presents a two- level safety control strategy, which balances the time cost for planning and safety for tracking. The top level obtains a preliminary trajectory through time efficient optimization but no safety is completely guaranteed. The bottom level formulates a Quadratic Programming problem incorporating Control Barrier Functions(CBFs) constraints to adapt the trajectory thus guaranteeing safety. Safety control is further improved by terrain map estimation with Fourier series and height compensation with empirical functions. The former smooths the sharp edges to satisfy CBFs' condition and the latter delivers a longer horizon to mitigate boundary riding dilemma. In the experiment section, our two-level safety control method is validated on Qingzhui hexapod robot in stairs climbing scenario.