Reconstructed informed RRT*algorithm for robotic manipulator trajectory planning

Abstract

Abstract Sampling-based path planning algorithms have demonstrated efficiency and been extensively applied in robotic arm end-effector trajectory planning. Bidirectional search-based strategies exhibit superior convergence and optimization potential. However, as dimensionality and spatial complexity increase, the computational time escalates significantly due to the algorithm’s reliance on global sampling and simultaneous optimization during sampling. This paper proposes Reconstructed Informed RRT* (RI-RRT*), which adopts RRT-Connect to generate initial trajectories and subsequently refines the random tree by pruning suboptimal nodes. The algorithm further integrates Informed-RRT*’s heuristic sampling strategy to enhance path optimization iteratively. This algorithm solves the disadvantages of long initial trajectory generation time and slow convergence of the bidirectional optimized random algorithm. Experimental validation in a 2D randomized environment demonstrates that the algorithm reduces path computation time by 28.9% compared to Informed-RRT* while achieving faster convergence than Bi-RRT* and RRT*-connect. Implementation in VREP simulation for UR5 robotic arm obstacle avoidance confirms its practical feasibility in high-dimensiona.