학술논문
A Chronological and Cooperative Route Optimization Method for Heterogeneous Vehicle Routing Problem
Document Type
Conference
Source
2023 Congress in Computer Science, Computer Engineering, & Applied Computing (CSCE) CSCE Computer Science, Computer Engineering, & Applied Computing (CSCE), 2023 Congress in. :68-75 Jul, 2023
Subject
Language
Abstract
This paper focuses on a deep reinforcement learning (DRL)-based approach for Heterogeneous Vehicle Routing Problem, where each vehicle in the fleet is characterized by its capacity and speed. Previous methods fail to optimize the route for a vehicle while cooperatively considering the current state (e.g., the remaining capacities and the positions) of other vehicles in the fleet. To solve this problem, we first propose a chronological Markov Decision Process, which makes the current state of the whole fleet available. Second, we propose a fleet encoder, a specific network architecture to promote cooperative route generation among vehicles by incorporating the fleet state. Experimental results show that our method outperforms other DRL-based methods on problems with any number of customers and vehicles.