This paper presents a sequential Model Predictive Control (MPC) approach to reactive motion planning for bipedal robots in dynamic environments. The approach relies on a sequential polytopic decomposition of the free space, which provides an ordered collection of mutually intersecting obstacle free polytopes and waypoints. These are subsequently used to define a corresponding sequence of MPC programs that drive the system to a goal location avoiding static and moving obstacles. This way, the planner focuses on the free space in the vicinity of the robot, thus alleviating the need to consider all the obstacles simultaneously and reducing computational time. We verify the efficacy of our approach in high-fidelity simulations with the bipedal robot Digit, demonstrating robust reactive planning in the presence of static and moving obstacles.
翻译:本文件介绍了在动态环境中对双翼机器人进行反应运动规划的顺序模型预测控制(MPC)方法。该方法依赖于自由空间的顺序多位分解,它提供一系列相互交叉障碍的顺序组合,自由多面和路径点。随后,它们被用来确定一个相应的多面控制程序序列,将系统推向一个避免静态和移动障碍的目标位置。这样,规划员将重点放在机器人附近的自由空间上,从而减少了同时考虑所有障碍和减少计算时间的需要。我们核查了我们在与双面机器人Digit进行高度纤维化模拟的方法的功效,在存在静态和移动障碍的情况下展示了强有力的反应规划。