Objetive A path planning method for unmanned farm machinery transfer is proposed to address the problems of relying on manual driving or manual management and planning, which are time-consuming, labor-intensive, and do not meet the requirements of unmanned applications.

Method ArcGIS was used to construct farm road layers and networks, and simulation experiments were conducted. A Dijkstra bidirectional search transfer path planning algorithm based on graph theory was developed, single and bidirectional searches were simulated using Python. A transfer path planning system based on a web platform was built.

Result In the simulation of the road network, the distances traveled by agricultural machinery from the hangar to the field, from the field to the field, and from the field to the hangar at a speed of 0.7 m/s were 241.57, 74.46 and 75.66 m respectively, with corresponding time of 345.10, 106.37 and 108.09 s. The single and bidirectional search time of Dijkstra’s algorithm were 0.632 and 0.216 s respectively, and the computational efficiency of bidirectional search was improved by 65.82% compared to unidirectional search. The transfer path planning system for agricultural machinery based on the web platform conducted real vehicle road tests at a speed of 0.7 m/s from the hangar to the field, from the field to the field, and from the field to the hangar. The absolute arithmetic mean difference between the sampling points of the path and the actual path of the agricultural machinery was less than 0.1 m, which met the transfer requirements of unmanned farm agricultural machinery. Compared to manual marking, the path planning efficiency of the transfer path planning system was higher.

Conclusion The constructed farm road layer, road network and transfer path planning system meet the road transfer needs of unmanned farm machinery. The research results can provide the technical support for the transfer path of agricultural machinery in unmanned farms.