To address the issues of slow convergence and excessive redundant nodes in traditional RRT-based algorithms for path planning, as well as the tendency of the artificial potential field (APF) method to fall into local optima, this paper proposes an adaptive RRT* path planning algorithm fused with an improved APF approach. The algorithm enhances the traditional APF by introducing a tangential force component from obstacles and employing an adaptive dynamic adjustment strategy, which dynamically modulates the magnitude of the tangential force based on path expansion difficulty, thereby improving the algorithm’s ability to escape local optima. A three-mode mechanism for expanding the random search tree is designed, in which the tree performs APF-guided expansion with probability p1, RRT* random sampling expansion with probability p2, and goal-biased expansion with probability p3. The probabilities of each mode are dynamically adjusted by statistically analyzing the success rates of tree expansions within a specified window, ensuring efficient convergence in complex environments. Additionally, a path node optimization strategy is incorporated to obtain a relatively optimal path. Simulation results demonstrate that the proposed algorithm significantly outperforms RRT-based and APF-based algorithms in simple environments. In complex environments, the average search time, path length, and number of turns are 62.7%, 71.8%, and 10.5% of those for the RRT* algorithm, respectively. The algorithm maintains the same time complexity as RRT*-based methods, with a significantly reduced standard deviation in the searched path lengths and a long-tailed distribution, indicating high efficiency, stability, and accessibility to optimal paths.