This paper proposes a new navigation method that makes a mobile robot move in dynamic unknown environments. Past dynamic motion planning methods are applied to low pedestrian density. If the robot is surrounded by a lot of pedestrians, it becomes difficult that the robot breaks away from the crowd. To overcome this problem, we introduce a new potential field method that makes the robot follow and avoid pedestrians without map information. The robot can move smoothly by joining a pedestrian flow that is headed in a leading vector. The leading vector indicates about the direction of the destination. In addition, the leading vector is updated by local path information, such as an informatory sign in a station, and pedestrians. Simulation results show the effectiveness of the proposed method.