In this work, we present a workspace-based planning framework, which though using redundant workspace key-points to represent robot states, can take advantage of the interpretable geometric information to derive good quality collision-free paths for even complex robots. Using workspace geometries, we first find collision-free piece-wise linear paths for each key point so that at the endpoints of each segment, the distance constraints are satisfied among the key points. Using these piece-wise linear paths as initial conditions, we can perform optimization steps to quickly find paths that satisfy various constraints and piece together all segments to obtain a valid path. We show that these adjusted paths are unlikely to create a collision, and the proposed approach is fast and can produce good quality results.
翻译:在这项工作中,我们提出了一个基于工作空间的规划框架,该框架虽然使用冗余的工作空间关键点代表机器人国家,但能够利用可解释的几何信息为即使是复杂的机器人提供高质量的无碰撞路径。我们首先为每个关键点找到无碰撞的无片段线性路径,这样每个关键点的端点都能够满足距离限制。用这些小片线性路径作为初始条件,我们可以采取优化步骤,迅速找到能够满足各种限制和将所有部分拼凑在一起以获得有效路径的路径。我们表明,这些经调整的路径不可能产生碰撞,而拟议的方法是快速的,能够产生高质量的结果。