@article{MA398DD65,
title = "Integration of MAML and PPO for Optimized Vehicle/Cargo Dispatch in Dynamic Environments",
journal = "The Journal of Korean Institute of Communications and Information Sciences",
year = "2025",
issn = "1226-4717",
doi = "10.7840/kics.2025.50.10.1515",
author = "Ji-Hyeon Kim, Soon-Young Kwon, Da-Min Shin, Hyoung-Nam Kim",
keywords = "Vehicle/cargo dispatch system, Transportation matching algorithm, MAML, PPO",
abstract = "For efficient vehicle and cargo dispatch in port logistics environments, it is essential to consider various environmental factors such as dispatch fairness/consistency, road conditions. These factors change dynamically in real-time and can significantly impact transportation vehicles and drivers. However, conventional optimization methods struggle to adapt to these environmental changes, making real-time dispatch optimization challenging.
In this paper, we propose a dispatch optimization approach that integrates model-agnostic meta-learning(MAML) and proximal policy optimization(PPO). MAML is employed to learn an optimized initial policy across diverse dispatch environments, and PPO is used to adapt to environmental changes and continuously refine the policy, thereby maximizing dispatch performance. Through simulations, we demonstrate that the MAML+PPO-based optimization algorithm outperforms traditional methods in dispatch performance. In particular, the aggregate performance measure(APM)-based performance analysis confirms that the proposed algorithm achieves high generalization performance and stable optimization across various scenarios, making it a robust solution for adaptive dispatch optimization in dynamic environments."
}