This article presents a novel method—Dynamic Environment Rapid Search Tree—for power transmission line maintenance robot, which is based on the learned field function of reachability and the genetic best-first policy. Dynamic Environment Rapid Search Tree uses a priori information to optimize the node selection strategy in the rigid–flexible coupling environment with slight perturbation and generates the joint path in the configuration space. While the search tree rapidly extends toward the goal configuration, it effectively avoids the obstacles and greatly reduces the expansion of the irrelevant region. Finally, the joint trajectory considering the dynamic constraints and cost function is given, which provides the reference positions and torques for the robot controller. Traditional planning algorithms are compared with our proposed method under two different operation modes, and the planner is demonstrated on the robot under real settings. The experiment results verify the feasibility and adaptiveness of the proposed algorithm and planner, even in slightly and continuously varying environment.

本文提出了一种新方法-动态环境快速搜索树-用于输电线路维护机器人，该方法基于学习的可达性字段函数和遗传最佳优先策略。动态环境快速搜索树使用先验信息来优化刚柔耦合环境中的节点选择策略，并产生轻微的扰动，并在配置空间中生成关节路径。搜索树在迅速向目标配置扩展的同时，有效地避开了障碍，并大大减少了无关区域的扩展。最后，给出了考虑动态约束和成本函数的关节轨迹，为机器人控制器提供了参考位置和扭矩。在两种不同的操作模式下，将传统的规划算法与我们提出的方法进行了比较，并在实际设置下在机器人上演示了规划器。实验结果验证了所提算法和计划器的可行性和自适应性，即使在持续不断变化的环境中也是如此。