Objective To solve the problem of traversal path planning of agricultural robot in field operation.Method A memory simulated annealing algorithm combined with A* algorithm was proposed. Firstly, the optimal walking sequence of target points in task was found by memory simulated annealing algorithm, and then A* algorithm was used for crossing regional linking of path planning.Result The simulation experiments showed that the Manhattan distance of traversal path planned by this algorithm was reduced by 9.4% compared with the traditional simulated annealing algorithm, the coverage of traversal path could reach 100%, and the repetition rate could be controlled at 4.2%.Conclusion Memory simulated annealing algorithm enhances the ability to jump out of the local optimal trap, and improves the quality of the solution obtained by adding memory device to the traditional simulated annealing algorithm. The research results can provide a theoretical basis for the path planning of agricultural robot.