Complete Coverage Path Planning for Agricultural Sowing Machine Using Improved Biological Neural Network

Complete coverage path planning for sowing machinery is a challenging task since sown areas should not be repeatedly traversed, which is a major difference with the planning of cleaning robots. This paper proposes an improved biological neural network (BNN) approach considering an operational mode switching mechanism between sowing and non-sowing movement. Based on surrounding environmental conditions, the next node state is classified as sowing, closed, or transfer node. Guided by the BNN landscape, the machine continues the sowing operation along parallel straight paths at sowing nodes. Once a closed node is detected, the machine switches to non-sowing mode and searches the potentially closed area using a depth-first search algorithm. Then, the machine moves to a new target node along the shortest non-sowing path and restarts sowing. This avoids the sown-path-induced enclosure (SPIE) problem. When a transfer node is detected, the machine also switches to non-sowing mode, then searches and travels to a reasonable new target node to restart sowing. This improves the rationality of the sowing path. Simulations show that the proposed method achieves complete coverage of sowing operations while avoiding repeated traversal of sown areas. Experiments on a differential drive crawler chassis platform verify the feasibility of the proposed method.