Objective To improve the path planning efficiency of multi-station food sorting robots.Methods A two-stage path planning method based on the improved gold rush optimizer (IGRO) was proposed. First, a two-layer planning model was constructed, incorporating the optimal grasping position at a single station and the shortest moving distance among multiple stations, with comprehensive consideration of three objectives: path length, smoothness, and safety. Second, the IGRO was designed by introducing free exploration behavior to enhance global search capability and adopting a cosine-shaped convergence factor to balance the exploration and exploitation efficiency of the algorithm.Results Compared with algorithms such as CRO and the ant colony algorithm, IGRO shortened the total path by 7.3%~39.2%, reduced running time by 26.7%~50.1%, and improved path smoothness by 18.6%.Conclusion The proposed method effectively improves the operational efficiency of food sorting robots and has high engineering application value.