Abstract: Intelligent agricultural robot is an important link to realize the complete replacement of manual labor by machines in unmanned farms, and path planning is an important key technology of agricultural robots. According to the special working environment of farmland and agronomic requirements, this paper proposes a global path planning method and a local optimal planning method for agricultural robots. The idea of full coverage path planning of farmland is expounded, and the full coverage path planning method based on neural network and the local optimal path planning method based on Dijkstra algorithm are designed. The feasibility of the path planning method is verified by MATLAB simulation test. The grid map is established, and the obstacle information is randomly set to simulate the actual working environment. The results show that the coverage completion rate of the two planning methods of biological excitation and neuron excitation is 100%, and the difference between the path length and the number of turns is small, However, the path repetition rate of the neuron excitation method is 5.49%,which is far less than that of the biological excitation method, which is 12.66%.The full coverage path planning method based on the neuron excitation network method has a lower re tillage rate than the full coverage path planning method based on the biological excitation network method, and can realize the large-scale coverage operation of farmland plots; The local optimal path planning method based on Dijkstra algorithm can realize autonomous optimal path planning for obstacle maps with different complexity.