Abstract: Aiming at the problems of slow convergence, tortuous paths, and susceptibility to local optima in the path planning of coal mine robots caused by unstructured environments and dynamic obstacle interference in underground coal mines, a path planning method integrating Artificial Potential Field (APF) and Multi-strategy Improved Dung Beetle Optimization (AIDBO) is proposed. Firstly, the Singer chaotic map is introduced to initialize the population, thereby improving the distribution uniformity of initial solutions in complex roadway environments. Secondly, the logarithmic spiral opposition-based learning strategy and non-linear adaptive constriction factor are fused to enhance the algorithm's ability to escape from local optima and synergistically balance global exploration and local exploitation. Performance tests on the CEC2005 benchmark function set demonstrate that the Improved Dung Beetle Optimization (IDBO) algorithm is superior to the standard DBO algorithm in terms of global search capability and convergence accuracy. Finally, a fitness-obstacle dual-driven gain adaptive strategy is constructed to embed APF into the algorithm as a post-processing operator, utilizing the potential field mechanism to perform secondary smoothing and local fine-tuning on the global path. Simulation verifications in grid maps of various dimensions show that the AIDBO algorithm exhibits significant advantages in both static and dynamic environments. In the 30×30 static scenario, the shortest path length of the AIDBO algorithm is 45.0537 cm, which is shortened by 15.4% and 22.0% compared with IDBO and DBO algorithms, respectively. In the 40×40 dynamic complex environment, the path length of AIDBO is 58.5261 cm, shortened by 20.3% and 28.2% compared with IDBO and DBO, respectively. Furthermore, after introducing the fitness-obstacle dual-driven strategy, the number of path inflection points is reduced by 15.38%, the cumulative turning angle is reduced by 21.77%, and the minimum safety distance is increased by 14.94%. The AIDBO algorithm not only significantly improves the convergence accuracy and optimization efficiency of path planning but also effectively solves the problems of path smoothing and safe obstacle avoidance in dynamic environments, demonstrating strong robustness.