Multi-Objective Reinforcement Learning for Supply Chain Optimization in Indonesia’s Free Nutritious Meal Program

Abstract

Indonesia’s Free Nutritious Meal Program (MBG) requires an efficient and adaptive supply chain system to ensure timely distribution, cost efficiency, and adequate nutritional delivery for a large number of beneficiaries. However, conventional supply chain approaches are generally static and unable to respond effectively to dynamic demand, supply uncertainty, and logistical constraints. This study proposes a Multi-Objective Reinforcement Learning (MORL) model to optimize the MBG supply chain by simultaneously considering distribution cost, delivery timeliness, service level, nutritional adequacy, and food waste reduction. The model is developed using a simulation-based environment representing real-world supply chain conditions, including demand variability, transportation limitations, and kitchen capacity constraints. The results show that the proposed approach achieves cost reductions of 15–22%, improves delivery timeliness by 18–25%, maintains a service level above 90%, increases nutritional fulfillment by 12–18%, and reduces food waste by 10–15% compared to baseline methods. Sensitivity analysis further demonstrates the robustness of the model, with minimal performance degradation under disruption scenarios. These findings indicate that Reinforcement Learning provides a scalable and adaptive solution for optimizing large-scale public food distribution systems. The proposed model contributes both theoretically by integrating multi-objective optimization within an RL framework and practically by supporting data-driven decision-making for improving the effectiveness of the MBG program in Indonesia.