A directed D* route planning algorithm was proposed, aiming at the low efficiency and high cost of the traditional D* route planning algorithms. In the proposed directed D* algorithm, the key nodes are defined by utilizing the location information of target points and obstacles so that the feasible routes could be determined by a stepwise expanding way, and a directed function is introduced to control the single searching range of nodes in order to improve the searching efficiency. A path smoothness function which punishes the deviation of paths is introduced to the algorithm in addition to Euclidean evaluation function, in order to avoid the redundant turning of robots and reduce the costs. The length of the path and the smoothness are taken into account simultaneously by the turning factor of the path smoothness function. The piecewise principle of the path smoothness function and the determining method of the turning factor are proposed. The convergence of the algorithm was also proved. Simulation experiments in different environments show that the proposed algorithm can balance the local searching and the global optimality, and it is especially suitable for complex environments with many obstacles.