A dynamic path planning is proposed for partly unknown environment. With the globally optimal path treated as a dynamically changed state, the dynamic path planning is able to be executed online by tracking the globally optimal path using particle filter. The tracking including a prediction from the motion of the robot and a updating from the up to date environmental information. So that dynamic path planning abandons the “following after planning” strategy which is generally adopted by global path planning approaches and adopts a “following while planning” strategy instead. Simulations and experiments show that, compared with the classical global path planning approaches, the proposed method improves the efficiency by reducing the time in waiting for planning result and provides a global adaptability to both the motion error of robot and the partly unknown environment.