Abstract: To solve the unsafe obstacle avoidance and low planning efficiency problems resulted from the more idealized obstacle model of the current underwater obstacle avoidance route planning algorithm, an autonomous real-time route planning algorithm based on the sub-object method and level set method was proposed. Based on the partial contour information of obstacles detected by AUV(autonomous underwater vehicle) forward-looking sonar, the algorithm was arranged to estimate the size and center position of obstacles in stages, and then to obtain safe and reliable sub target points. And then, a rolling sub target method was used to realize complete obstacle avoidance and a level set method was used to improve the planning speed. The simulation results show that the novel algorithm can achieve 100% obstacle avoidance, and the performance quality of the route after obstacle avoidance appears a little debased than that of the global optimal route obtained by the level set method:the average increase of route length and sailing time are less than 5%.