Graph Attention Network–based Delivery Route Generation Incorporating Human Experiential Data

Key Points

• The research aims to create an efficient delivery route generation method that incorporates human experiential data.
• Combined Graph Attention Network (GAT) with Pointer Network for route generation.
• Utilized actual delivery driver route data for modeling.
• Implemented location clustering and movement frequency information.
• Applied a 2-opt based route improvement technique.
• The proposed method maintained structural consistency in delivery routes.
• Showed similar travel times to real delivery routes.
• Indicated a successful integration of human experience in route planning.

Abstract