MI-RRT*: an improved RRT* algorithm for mobile robot path planning with multi-informed sampling and two path optimization strategies

The Rapidly exploring Random Tree (RRT) method has been extensively utilized in robotic path planning over the past decade. Among its variants, the RRT* algorithm is considered the most classical, as it enhances path quality while guaranteeing asymptotic optimality. However, further improvements in path quality and computational efficiency are still required for RRT*. To address these issues, an enhanced RRT* algorithm, named multi-informed RRT* (MI-RRT*), is proposed in this paper, with modifications focused on path optimization and sampling strategy. Regarding sampling strategy, the multi-informed sampling was proposed, where the sampling region is adaptively narrowed during iterations to accelerate convergence. This strategy can also be extended to high-dimensional spaces. For path optimization, the Bid-backtracking pruning and the Fast-optimization strategies are introduced. Both strategies are demonstrated to reduce path cost effectively, and their applicability is confirmed for general path optimization problems. To evaluate the algorithm’s performance, three simulation environments were designed, and comparative experiments with existing methods were conducted. The results show that MI-RRT* achieves a significant reduction in computation time while generating higher-quality paths across all tested scenarios, thereby successfully balancing efficiency and optimality.