A city logistics transportation optimization solving method based on spatial distance attention

CN122243340APending Publication Date: 2026-06-19BEIJING INST OF TECH

Patent Information

Authority / Receiving Office: CN · China
Patent Type: Applications(China)
Current Assignee / Owner: BEIJING INST OF TECH
Filing Date: 2026-03-19
Publication Date: 2026-06-19

AI Technical Summary

Technical Problem

Existing vehicle routing problems (VRP) struggle to achieve efficient and rapid optimal route planning in complex traffic environments in urban logistics transportation, especially in uncertain scenarios such as new order arrivals and sudden traffic changes, where real-time rerouting is difficult.

Method used

We employ a deep reinforcement learning approach based on spatial distance attention. By using the spatial distance attention mechanism in the encoder and decoder of the Transformer network, we capture the clustering and dispersion characteristics of customer nodes. Combined with the urban road network structure, we train the model to select the optimal route and use reinforcement learning for path optimization.

Benefits of technology

It achieves high-quality optimal path solutions for urban logistics transportation within milliseconds, outperforming heuristic algorithms, and maintains high efficiency even with a large number of users.

✦ Generated by Eureka AI based on patent content.

Smart Images

Figure CN122243340A_ABST

Patent Text Reader

Abstract

This invention addresses the field of vehicle routing problems, specifically involving a method for optimizing urban logistics transportation based on spatial distance attention. The method comprises the following steps: (1) establishing a CVRP objective function under urban logistics transportation conditions, constructing a mathematical model of the CVRP problem, and defining the constraints of the model; (2) constructing a reinforcement learning environment for the urban logistics transportation optimization problem; (3) building a deep learning network based on a spatial distance attention mechanism according to the reinforcement learning environment; and (4) training the constructed deep learning network using a real-world vehicle routing problem dataset containing traffic data to obtain an application model that meets pre-defined requirements, and verifying the model's application in specific cases. The deep reinforcement learning algorithm designed in this invention demonstrates good solution quality and speed on urban logistics transportation problem instances of different scales after training.

Need to check novelty before this filing date? Find Prior Art