Variable Step Size Strategy for RRT* Algorithm

Pathfinding algorithm play a crucial role in the field of mobile robots. Among various algorithms, RRT* stands out as a representative sample-based approach that is increasingly utilized in complex environments due to its computational efficiency and minimal reliance on obstacle map information. However, the key to RRT*’s effectiveness lies in its convergence rate, given its asymptotic optimality. To address this challenge, this paper presents a novel Variable Step Size (VSS) strategy based on RRT*. The VSS strategy dynamically adjusts the expansion step size based on both the direction of the vertex and the goal point in the random tree, aiming to reach the goal point more rapidly. Since various variants of RRT* already involve extended steps, the VSS strategy exhibits excellent applicability in practice. Furthermore, VSS significantly enhances the likelihood of connecting the random tree to the goal point, facilitating faster identification of the initial path to initiate the optimization phase. Leveraging the optimization characteristics of VSS, when combined with the optimization methods employed in different variants of RRT*, the convergence rate of the algorithm can be further accelerated. In the simulation results, VSS combines well with the RRT*, RRT*-Connect, Informed- RRT*, Improved- RRT* and RRT*-Smart, not only reducing the number of iterations of the initial path, but also speeding up the convergence.