Adaptive Goal-Biased Bi-RRT for Online Path Planning of Robotic Manipulators

Autonomous obstacle avoidance path planning plays a crucial role in enabling intelligent and safe operation of robotic manipulators. While the RRT algorithm exhibits promising performance in high-dimensional spaces by avoiding explicit environment modeling and exploring a wide search space, its efficiency is obstructed by random and blind sampling and expansion, limiting online planning and efficient motion capabilities. To address this, we propose the Adaptive Goal-Biased Bi-RRT (AGBBi-RRT) algorithm, which incorporates a goal-biased strategy and an adaptive step size strategy to enhance path planning efficiency in high-dimensional spaces. By combining the concept of Artificial Potential Fields (APF) with a goal-bias factor, we guide the expansion of tree nodes, mitigating the limitations of the RRT algorithm. Additionally, to overcome falling into ‘local minimum’, an adaptive step size strategy is proposed to enhance obstacle avoidance capabilities. Further improvements are achieved through bidirectional pruning and cubic non-uniform B-spline fitting, resulting in shorter and smoother paths. Simulation experiments are conducted to evaluate the performance of AGBBi-RRT algorithm on a six-dof robotic manipulator in single-obstacle and multi-obstacles scenarios, comparing three algorithms: Bi-RRT, GBBi-RRT, and AGBBi-RRT. The experimental results show significant improvements in AGBBi-RRT compared to the previous two algorithms. In the single-obstacle scenario, the AGBBi-RRT algorithm achieves a 15.28% and 9.11% reduction in calculation time, while in the multi-obstacles scenario, the reduction is 27.00% and 15.92% respectively. In addition, AGBBi-RRT exhibits encouraging performance in terms of number of nodes, number of waypoints, and path length.