面向全方位双足步行跟随的路径规划

双足步行机器人的足迹规划方法难以满足快速步行条件下的计算效率要求, 并存在步幅变化时运动失稳的风险, 2D环境下点机器人栅格规划则难于生成针对双足步行的高效路径.本文提出针对各向异性特征全方位步行机器人的一种路径规划策略, 将状态网格图方法拓展到全方位移动机器人领域, 基于三项基本假设及基元类型划分给出了系统的运动基元枚举及选择方法, 借助实时修正的增量式AD*搜索算法实现仿人机器人在动态环境下的快速路径规划, 通过合理选择启发函数及状态转移代价, 生成了平滑高效的路径, 为后续足迹生成的动力学优化提供了基础.计算机仿真证实了方法对各类环境的适应性, Robocup避障竞速挑战赛的成功表现证明了方法对于机器人样机部署的可行性及其提高步行效率的潜力.

Omni-directional Bipedal Walking Path Planning

Footstep planning for bipedal walking robot is difficult to fulfil the requirement of computational efficiency at high walking speed; it also suffers from the risk of falling over with stride variations. On the other hand, 2D grid-based planning strategy for point robots is unable to generate an efficient walking path for bipedal robots. A path planning approach for heterogeneous omnidirectional bipedal walking robot is proposed in this paper. State lattice graph is brought into the omnidirectional moving condition. Based on three assumptions and type classification, a systematic motion primitive enumerating and selecting method is given. The rapid path planning for humanoid robots in dynamic environment can be achieved by making use of the algorithm of anytime repairing incremental search AD*. The smooth and efficient path, generated by reasonable heuristics function and state transfer cost, provides the foundation for dynamics optimized footstep planning. Simulation has demonstrated the adaptability of the biped robot in various environments. The successful performance in obstacle avoidance challenge in Robocup proves the possibility of implementation on physical robots and the capacity for walking efficiency improvement.