关键词: 农业机器人, 路径规划, 蚁群算法, 差分 Transformer, 智慧农业

Abstract: To address the challenges of low path planning accuracy and limited obstacle avoidance capabilities of agricultural robots operating in complex field environments, this study proposes a novel path planning method that combines the Ant Colony Algorithm (ACA) with a differential Transformer. Initially, the ACA is for global path search, leveraging its distributed and parallel search capabilities to generate an initial feasible path. To overcome the traditional ACAs limitations, such as susceptibility to local optimization and poor adaptability to dynamic changes of the environment, a differential Transformer model was introduced to replace the conventional pheromone updating mechanism. By utilizing a self-attention mechanism, the differential Transformer captures long-range dependencies and nonlinear features between path nodes, thereby allowing for more precise pheromone updates and better adaptability in complex conditions. Experimental results showed that the proposed method outperforms traditional algorithms in terms of path length, planning time, and obstacle avoidance success rate. Specifically, in an environment with a grid size of 50, the average path length was reduced by 16.8%, from 150 meters to 125 meters. Planning time was shortened by 23.5%, from 2.13 seconds to 1.63 seconds.The obstacle avoidance success rate increased by 11.2%, reaching 96.5%. This research provides an effective solution for autonomous navigation in agricultural robotics and holds significant theoretical and practical value.

Key words: agricultural robot, path planning, ant colony algorithm, differential Transformer, smart agriculture