UAV path planning based on a multi-strategy adaptive rapidly-exploring random tree star algorithm

In recent years, the Rapidly-Exploring Random Tree Star (RRT*) algorithm has drawn significant attention in academic circles owing to its properties such as asymptotic optimality and high exploration efficiency. Nevertheless, it still suffers from the drawback of a relatively high degree of randomness in the search path. Therefore, this paper presents a multi-strategy adaptive RRT* algorithm, termed MARS-RRT*. First, a target-biased strategy is introduced to guide the tree expansion toward the goal region, thereby reducing unguided exploration. Secondly, an improved version of the Artificial Potential Field (APF) algorithm is incorporated into the Random Tree. Additionally, rotational Probing and Random Point Direction are employed to help the algorithm break away from local optimal traps. Finally, Triple-Stage Path Smoothing is implemented to guarantee the optimality of the path. This paper compares this algorithm with RRT*, GB-RRT*, APF-RRT*, BI-RRT*, BI-APF-RRT*, and IBI-P-RRT* algorithms. The proposed method is validated through comparative simulations in five three-dimensional environments. Experimental results against RRT*, GB-RRT*, APF-RRT*, BI-RRT*, BI-APF-RRT*, and IBI-P-RRT* demonstrate that the proposed algorithm generates shorter trajectories with fewer redundant nodes, converges in fewer iterations, requires less runtime, maintains larger minimum obstacle clearance, and produces smaller average turning angles.