Path planning for robotic arms using an improved Q-RRT* algorithm in a digital twin system

Abstract

This study addresses the shortcomings of the traditional RRT* algorithm in path planning, such as high path costs and slow convergence speeds, and proposes an improved algorithm with dual-tree collaborative characteristics—KQ-RRT*-Connect. The algorithm builds upon the Q-RRT* framework by designing a dual random tree target-biased sampling strategy, dynamically adjusting the sampling probability density, employing a KD-tree spatial index structure to optimize nearest neighbor queries, introducing a bidirectional midpoint optimization model, and combining a second-order Bessel curve to optimize paths. Through these multi-optimization mechanisms, the algorithm simultaneously enhances the efficiency and quality of path planning. The algorithm has been validated through two-dimensional and three-dimensional environment simulations in MATLAB and further experimentally verified on a masonry robot digital twin platform and a UR10 robotic arm system. Experimental results show that the algorithm effectively suppresses exploration of invalid regions, reducing the time complexity of neighborhood search from O ( n ) to O (log n ), and improves path continuity by reducing redundant nodes; The proposed modifications reduced the initial path calculation time of O-RRT* by 90%, reduced the number of iterations required by 85%, and reduced the path cost by more than 4.5%. It also demonstrates strong robustness and promising engineering application prospects.