An improved genetic algorithm is proposed for the path planning of a bulk solids clearance robot using a terrain traversability raster map. The traversability raster map is established by a fuzzy logic inference method based on the environmental terrain characteristics and robot traversability. A new fitness function involving the path length, smoothness and the traversability cost, is also developed. Logistic chaos mapping is used for initializing the population to enhance the randomness of the initial population distribution, and a catastrophic population update mechanism is introduced to improve the genetic algorithm. The simulation results show that the paths planned based on the improved genetic algorithm have higher quality and faster convergence compared to the original algorithm.