Robot express sorting model based on feature enhancement network

Robot express sorting system has become a crucial research topic in express logistics. To optimize the effectiveness of sorting robots in the field of express logistics, an express object detection model and a sorting robot path planning model are proposed by introducing a spatial pyramid pooling structure, a depthwise separable convolution, and a potential energy reward function. The average precision for detecting express delivery was 97.55%, the average recall was 97.71%, the floating point operations was 18.0G, and the forward pass latency was 9.44ms. The collision rate on a simple map was 2.45%, and the collision frequency was 24 times. The collision rate on complex maps was 26.74%, with a collision frequency of 98. In summary, the detection model has good detection performance and excellent detection speed. The proposed method performs well in simple and complex maps. The express sorting model has effectively improved the operational efficiency and promoted the development of the express logistics field.