Optimizing Fairness in Production Planning: A Human-Centric Approach to Machine and Workforce Allocation
This paper addresses the joint optimization of operational efficiency and workforce fairness in industrial production planning. We propose a bi-level intelligent scheduling framework: an upper-level constraint programming (CP) model optimizes order–production-line assignment while satisfying hard constraints—including machine capacity, processing time, and due dates; a lower-level Markov decision process (MDP) incorporates worker preferences, experience, adaptability, and medical restrictions to dynamically assign tasks, with comparative evaluation of greedy heuristics, Monte Carlo tree search (MCTS), and reinforcement learning (RL). Our key contribution lies in tightly coupling deterministic optimization with learning-based decision-making to achieve human-centered fair scheduling without compromising throughput or latency. Experiments demonstrate statistically significant improvements over baselines in fairness metrics (e.g., workload variance reduction) and preference satisfaction rate. The approach has been validated and highly endorsed by domain experts in automotive manufacturing.