Abstract: To address the problems that path planning algorithm for coal mine inspection robots based on Grey Wolf Optimization (GWO) is prone to falling into local optima and shows insufficient dynamic adaptability in complex underground environments, an Improved GWO (IGWO)-based path planning algorithm for coal mine inspection robots was proposed. A Piecewise Linear Chaotic Map (PWLCM) was introduced for population initialization to ensure uniform population distribution and enhance global search capability. A nonlinear convergence factor was designed to effectively balance the algorithm’s global exploration and local exploitation and avoid falling into local optima. A dual-population structure, differential evolution, and an elimination mechanism were introduced to enhance population diversity and improve the algorithm’s adaptability to the environment. A cubic B-spline curve was incorporated to smooth the generated path, improving path executability and reducing redundant turning points. A two-dimensional spatial model based on feature grids was proposed to effectively reduce the computational complexity of path search and improve real-time performance of the algorithm. Comparative experiments were conducted under various typical coal mine environments including random obstacle maps, fixed obstacle maps, and narrow maps, with IGWO compared against GWO, Memory, Evolutionary Operator, Local Search, and Linear Population Size Reduction Technique based GWO (MELGWO), A^*, Whale Optimization Algorithm (WOA), and Particle Swarm Optimization (PSO). The results showed that IGWO outperformed the comparison algorithms in terms of path length and safety. In random complex scenarios, the path length of IGWO was 56.9% shorter than that of MELGWO. In a 20×20 fixed scenario, the average number of turning points of IGWO was reduced by 12.5% and 44.4% compared with WOA and A^*, respectively. In a 40×40 fixed scenario, the range and variance of the IGWO path length were both lower than those of WOA and PSO. In a narrow map environment, IGWO successfully planned a smoother path than A^* and required less running time.