Dynamic Multi-Objective Decision Making in Supply Chain Management

This research article discusses the integration of reinforcement learning (RL) and multi-objective evolutionary algorithms (MOEAs) for dynamic multi-objective decision-making in supply chain management. The document outlines the limitations of traditional RL approaches, which typically focus on single objectives, and highlights the need for methods that can handle multiple conflicting non-differentiable objectives. The proposed framework employs a derivative-free approach to optimize these objectives effectively. The article presents two case studies demonstrating the adaptability of the proposed method in various scenarios. It emphasizes the importance of fast decision-making in supply chains, particularly in the face of disruptions, and the necessity of balancing profitability, social responsibility, and environmental sustainability. The methodology aims to provide a diverse set of policies that can help decision-makers navigate the complexities of modern supply chains, which are characterized by high interconnectivity and uncertainty. The paper is structured to introduce the preliminaries, methodology, problem statement, and case studies in a logical sequence.