Abstract: A hierarchical robust motion planning method is proposed to address the challenges of trajectory planning inefficiency and stability degradation arising from the intrinsic shape and anisotropic locomotion characteristics of quadruped robots in map-free complex environments. For target positioning in environments lacking complete maps, a polynomial MiniSnap algorithm combined with a weighted hybrid A* algorithm is employed at the front end to initialize the trajectory, effectively mitigating difficulties in trajectory tracking and reducing the tendency to become trapped in local optima. Meanwhile, to enhance both the efficiency and quality of local collision-free path searching, a weighting factor is integrated into the evaluation function, and constraints are imposed on yaw angle and lateral motion. Based on the initial trajectory obtained, the back-end adopts a strategy of optimizing B-spline control points by formulating an optimization model that considers constraints related to trajectory smoothness, collision avoidance, and dynamic feasibility, and iteratively refines the solution to provide a smooth, safe, and executable path for the quadruped robot. Experimental results show that the proposed method reduces planning time by 24% and execution time by 26.5% in obstacle-dense environments without prior maps, while demonstrating superior robustness compared to the mapless trajectory planning algorithm Ego-Planner.

Key words: quadruped robots, motion planning, trajectory optimization, mapless autonomous navigation

摘要： 针对四足机器人在无全局地图的复杂环境中因自身形状和各向异性运动特性导致的运动规划效率低和稳定性差等问题，提出了一种面向无地图环境下的四足机器人分层鲁棒运动规划方法。对于缺乏完整地图环境中定义的目标位置，前端采用多项式MiniSnap及加权混合A*算法初始化轨迹，解决轨迹难以跟踪和容易陷入局部最优等问题；同时，为了提高局部无碰路径搜索的效率和质量，提出在评价函数中引入权重因子，并约束偏航角和侧向运动；基于获取的初始轨迹，后端采用优化B样条曲线控制点的策略，考虑轨迹的平滑性、碰撞性、动力学可行性等约束目标构建优化模型，通过迭代优化为四足机器人提供一条平滑、安全的可执行路径。实验结果表明，所提方法在无地图障碍物密集的环境中规划时间降低了24%，执行时间减少了26.5%，相比于无地图轨迹规划算法Ego-Planner表现出更优的鲁棒性。

关键词: 四足机器人, 运动规划, 轨迹优化, 无地图自主导航