Efficient and accurate path planning algorithm of unmanned vehicles for unstructured road is of great significant for field disaster relief. An improved A–star algorithm for unmanned vehicle path planning is proposed in this paper to achieve optimized path with shorter length and less inflection point. Redundant safety spaces are set up in the improved algorithm to filter out impassable narrow roads and to avoid collisions between vehicles without obstacles. The prejudgment planning strategy and redundant inflection point elimination strategy are integrated into the algorithm to obtain a better driving path. Furthermore, a safe corridor approach was used in the trajectory optimization to avoid repeated checks and corrections and save computer time. The simulation results show that the path planned with the improved algorithm has shorter length, less inflection point and smoother than the path planned with the initial algorithm, and can avoid collision between vehicle contours and obstacles.