Abstract: The working efficiency of orchard robot in complex environment is mainly determined by the path planning of robot. There are many problems in the basic artificial potential field algorithm, such as deadlock, long running time and so on. Based on this, a new algorithm was proposed to improve the basic artificial potential field algorithm. The first is to reconstruct the recognition path, introduce the obstacle detection algorithm, which could identify the effective obstacles and calculate the effective middle point of the path. Then it was optimizd the repulsive force, introduced the boundary conditions related to the obstacles, and then mapped the information matrix to the matrix, calculated the direction of the maximum force. At the same time, re-searching from the starting point in this direction, and setting the intermediate point as the new starting point to iterate repeatedly, so that a local optimal path could be obtained, and the final global path was composed of each local optimal path. Through the comparative test of MATLAB, the test results showed that: compared with the basic artificial potential field algorithm, the path length of this algorithm was reduced by 0.586 meters, the number of iterations was reduced by 19, and the running time of the algorithm is reduced by 8.662 seconds. In the complex environment, the advantages of this algorithm are obvious.