For the problem of real-time path optimization for picking robot manipulator, an improved artificial potential field method was presented for apple picking robot manipulator to avoid obstacle under unstructured environment. According to the concrete structural characteristics of 5-DOF apple picking robot manipulator and obstacle, the manipulator model and obstacle model were analyzed. On the basis of retaining the merits of the traditional artificial potential field method which had simple structure and easy implement, for some of the shortcomings of its existence, such as local minimum points, the stuck district, combined with the characteristics of obstacles in the apple growing environment, a virtual target point was introduced to help the search process escape local optimal minimum point and obtain an optimization smooth motion curve. This method was applied in the obstacle avoidance path planning experiment for the apple picking robot manipulator, which included end-effector, obstacle and targets position. The experimental result showed that the proposed method was simple and enabled to overcome the shortcoming of traditional artificial potential field method.