An optimal and efficient hierarchical motion planner for industrial robots with complex constraints

摘要

This paper investigates the motion planning problem for industrial robots with complex constraints. An optimal and efficient hierarchical motion planner is proposed to obtain high-quality trajectories with low computational effort. First, the motion planning problem is formulated as an optimal control problem incorporating robot kinematics, obstacle-avoidance, and dynamics constraints. Thereafter, a hierarchical framework is constructed within the Markov decision process, consisting of two planners. In the high-level planner, a reinforcement learning-based policy is employed to generate virtual targets for the robot to navigate around obstacles, which can avoid collision detection. Then, in the low-level planner, a global orthogonal collocation method is used to generate time-energy optimal trajectories, considering dynamics, path, and boundary constraints such as joint position, velocity, and torque. Finally, simulation results on a 6-DOF (degree of freedom) and a 7-DOF industrial robot validate that the proposed method can produce high-quality trajectories with improved success rates and computation times compared to existing works.