Online step adaptation is critical for legged locomotion to resist disturbances, while current approaches mainly focus on step location regardless of step timing. This paper proposes an online adaptation algorithm that deals with both step timing and step location. A model predictive control framework is developed based on centroidal dynamics to predict robot state trajectories, which are utilized to simultaneously optimize step timing and location with full considerations on capturability, feasibility and reachability. Then, a swing foot trajectory generation method is presented to accommodate mutiple constraints on kinematics and actuators. Moreover, a heuristic compensation is enhanced to the step location for quadruped trotting locomotion, increasing the anti-disturbance ability of legged locomotion. Various push-recovery simulation and experiment on an 18 degree-of-freedom legged robot are implemented to verify the proposed algorithm.

在线步调自适应对于腿部运动抵抗干扰至关重要，而当前的方法主要关注步调位置，而与步调时机无关。本文提出了一种在线自适应算法，该算法同时处理步距计时和步距定位。基于质心动力学开发了模型预测控制框架，以预测机器人状态轨迹，该模型可在同时考虑可捕获性，可行性和可及性的同时，用于优化步距计时和位置。然后，提出了一种摆脚轨迹生成方法，以适应运动学和执行器上的多种约束。而且，对于四足小跑运动，启发式补偿被增强到台阶位置，从而增加了腿式运动的抗干扰能力。