Dynamic Spherical Sampling-Based Path Planning Method

Achieving efficient and reliable path planning in complex environments is a significant challenge for autonomous robotic arm operations. Although the traditional Rapidly-Exploring Random Tree (RRT) algorithm possesses probabilistic completeness, it is prone to issues such as high sampling redundancy, substantial computational resource consumption, and tortuous paths in environments with high obstacle density. This paper proposes an improved path planning method. By forming a spherical uniform probability distribution sampling around the target point, it reduces ineffective exploration. By adaptively adjusting the step length based on the perceived local obstacle density, it reduces computational resource consumption. Additionally, it incorporates path smoothness constraints into the real-time path extension in the form of an evaluation function to avoid excessively tortuous planned paths. Simulation experimental results show that, in two-dimensional and three-dimensional environments with the same obstacle density, the proposed method outperforms traditional algorithms in sampling efficiency, achieving a 25% improvement. Furthermore, when the obstacle density significantly increases, the planning success rate of traditional algorithms decreases by 40%, whereas the proposed method only experiences a 5% decrease, validating its stability and applicability in high obstacle density environments.