Research on pear picking path planning based on deep reinforcement learning

This study proposes RouteSolveformer, an enhanced Transformer-based model for pear picking path planning, leveraging deep reinforcement learning. The model uses an encoder-decoder structure tailored for path optimization and is trained by framing the task as a Markov Decision Process. Evaluation metrics include path length, deviation from the shortest path, and computation time. Experimental results demonstrate that RouteSolveformer achieves faster and more accurate path planning compared to traditional and heuristic methods. It performs exceptionally well in solving medium- and large-scale path planning problems, showcasing its effectiveness and potential for intelligent agricultural applications.