Abstract: An obstacle avoidance path planning method of manipulator based on proximity is proposed. Firstly, the concept of "movement direction" of link is defined for the obstacle perception problem based on sparse distance information of proximity sensor array. The sensors in the movement direction of link are read preferentially to reduce the read period. The nearest obstacle surface is modeled as the "imaginary cone", through whose vertex a "safety plane" is made. The link can't collide with the safety plane in the movement process. Secondly, an artificial potential field method based on potential function and joint space is applied to the obstacle avoidance path planning problem. According to the proposed obstacle perception method, an improved method based on "detour" is proposed to solve the local optimum problem of the artificial potential field method. Finally, the proposed method is verified on the forearm of a UR10 robot, and the experimental results show that the proposed method is effective, robust and real-time.