Abstract: Aiming at the problems of A* algorithm, such as a large number of redundant nodes, low smoothness, and a high tendency to fall into local minima, improvements are made by an adaptive variable-step-size search strategy to obtain a shorter, smoother, and safe and feasible mobile path in this paper. Firstly, the weight of the heuristic function is designed to ensurethat the intelligent agent can balance the relationship between the moving path and obstacles. Secondly, an adaptive variable-step-size search strategy is introduced to improve the algorithm's adaptability to complex maps. Finally, the Floyd pruning algorithm is introduced to further simplify the path and remove redundant turning nodes. Simulation results demonstrate that the improved A* algorithm has obvious advantages in terms of path length and smoothness. Compared with the traditional A* algorithm, the turning angle of the planned path is reduced by 64.69% and the path length is reduced by 12.9%. Furthermore, compared with the RRT* algorithm, the improved A* algorithm also has obvious advantages in terms of path length, smoothness, and computational efficiency.