Abstract: To address the problems of the basic Grey Wolf Optimizer (GWO), such as its tendency to fall into local optima and insufficient dynamic adaptability in path planning for the complex underground coal mine environment, this paper proposes an Improved Grey Wolf Optimizer (IGWO) for path planning of coal mine inspection robots.The algorithm employs a Piecewise Linear Chaotic Map (PWLCM) for population initialisation, enhancing the uniform coverage of the solution space in narrow roadways. A non-linear convergence factor is designed to balance global exploration and local exploitation capabilities. A dual-population structure and differential evolution strategy are introduced to enhance population diversity. Cubic B-spline curves are used to smooth the path, combined with a two-dimensional space modelling method based on feature grids, effectively reducing the complexity of path planning. Comparative simulation experiments of IGWO with basic GWO, MELGWO, A*, WOA and PSO algorithms are conducted in several typical coal mine environment models (including random obstacle maps, fixed obstacle maps and narrow mine roadway maps). The results show that the IGWO algorithm performs better than the comparison algorithms in terms of path length and safety. In random complex scenarios, the IGWO path length is reduced by 56.9% compared to MELGWO. In a 20×20 fixed scenario, the average number of inflection points of IGWO is reduced by 12.5% and 44.4% compared to WOA and A*, respectively. In a 40×40 fixed scenario, the range and variance of the IGWO path length are better than those of WOA and PSO. In a narrow mine roadway environment, IGWO successfully plans a smoother path than A*, with a shorter runtime.