Path planning is crucial for ensuring precise task execution and maintaining reliable robot operation. However, the fast marching method (FMM) faces challenges such as prolonged computational time, high costs, and limited adaptability in large-scale, complex environments. To address these issues, this paper proposes an adaptive scale hybrid fast marching method (AS-HFMM) that leverages adaptive map scales. In AS-HFMM, a quadtree-inspired approach is employed, which iteratively merges \(2 \times 2\) grid cells based on their obstacle density to reduce the map’s complexity. To preserve essential details, the process terminates when the difference in obstacle ratios between the scaled and original maps exceeds a predefined threshold. This effectively reduces the computational load for high-precision pixel-level grid maps. Using the simplified map, FMM performs preliminary path planning to generate an approximate route. This route is then projected onto the original map through scale-based coordinate transformation. To refine the path, a visibility-based method is adopted to prune redundant nodes, followed by local path smoothing using a segmented minimum-snap technique to enhance path smoothness. Comparative simulation results demonstrate that AS-HFMM reduces computational overhead by 5.59% and path length by 5.17%, outperforming the traditional FMM.
