Because there is some shortcomings in basic Rapidlyexploring Random Tree (RRT) algorithm such as a great quantity of branches, strong randomness, this paper presents an improved RRT. This paper mainly deals with the path planning of a series manipulator in threedimensional (3D) obstacle avoidance workspace. For avoiding obstacles acutely and effectively, the wall plane and Window function are introduced the Window RRT (shorted as Win_RRT) algorithm. The Win_RRT algorithm can not only combine with environmental constraints, but also generate the identical path for the same environment. The path is found by the Win_RRT algorithm in the obstacle avoidance workspace of Monte Carlo. Finally, the accuracy and effectiveness of the algorithm are proved by a manipulator example on MATLAB.