To address the issues of uniform repulsive influence range in traditional artificial potential field methods that lead to path redundancy, low planning efficiency, and susceptibility to local optima in complex multi-obstacle environments, an improved artificial potential field path planning method based on adaptive elliptical repulsive field is proposed. First, an adaptive elliptical repulsive field theory is constructed, utilizing LiDAR to obtain real-time geometric features of obstacles and dynamically adjusting the elliptical repulsive influence range based on multi-dimensional factors. Second, a virtual target point generation method based on dual-circle strategy is designed, determining the optimal escape direction through geometric construction to effectively solve the local optimum problem. Finally, the algorithm performance is validated through simulation experiments and mobile robot physical platform verification. Simulation results show that in complex discrete obstacle environments, compared with existing improved methods, the proposed algorithm reduces runtime by 48.19%, path length by 4.67%, and path deviation by 5.65%, with significantly lower escape mechanism trigger frequency. Physical experiment validation demonstrates that compared with existing improved methods, the proposed algorithm achieves optimization of 20.15%, 14.95%, and 5.95% in runtime, path length, and path deviation respectively. Experimental results prove that the improved algorithm exhibits excellent performance in path planning efficiency, safety, and environmental adaptability, with good robustness and real-time performance, capable of meeting the path planning requirements of mobile robots in complex multi-obstacle environments.