TRAJECTORY PLANNING OF ACUPUNCTURE ROBOTS

Abstract

For the clinical demand of sequential needling at multiple facial acupoints, needles already inserted and retained in the tissue become static obstacles to subsequent insertions. Planning a collision-free, smooth and efficient insertion path for each new needle inside the narrow facial workspace is therefore a key problem in the automation of acupuncture robots. This paper takes an acupuncture robot system composed of a UR5e six-degree-of-freedom manipulator and a self-developed acupuncture end effector as the research object, formulates multi-acupoint sequential needling as a constrained motion planning problem in which the obstacle set grows with the needling sequence, and proposes an improved APF-RRT* algorithm. The algorithm combines a bidirectional growth mechanism and goal-biased sampling on the RRT* framework to improve the convergence of the two trees, introduces Informed ellipsoidal subset sampling to reduce ineffective exploration, constructs an improved artificial potential field to alleviate the goal-unreachable and local-minimum problems, and designs a dynamic adaptive step-size strategy based on the ratio of attractive to repulsive force together with a bidirectional greedy path-pruning strategy. Simulation results in two- and three-dimensional environments show that, under the same maximum number of nodes, the proposed algorithm converges faster and yields a lower path cost than RRT* and APF-RRT*: in the 2D environment the runtime is reduced from 44.702 s to 9.538 s and the path cost from 792.702 to 750.477, while in the 3D environment the runtime is reduced from 84.594 s to 32.796 s and the path cost from 933.077 to 822.857, verifying the effectiveness and engineering applicability of the algorithm.