Optimizing Fairness in Production Planning: A Human-Centric Approach to Machine and Workforce Allocation

TL;DR

The paper tackles production planning in manufacturing by addressing both operational efficiency and workforce fairness. It proposes a two-layer approach that uses Constraint Programming to optimize Order–Line allocations and an MDP (with RL and MCTS options) to allocate workers to lines while accounting for worker preferences, experience, resilience, and medical constraints. Empirical evaluation with automotive-domain experts demonstrates strong performance: dense, high-utilization CP schedules with low tardiness and improved fairness from the MDP-based worker assignment, though expert feedback points to improvements in due-date adherence, setup integration, and assignment continuity. The work shows that combining CP with learning-based decision-making yields a robust, configurable foundation for fair and efficient production scheduling suitable for real-world deployment via a service-oriented architecture.

Abstract

This work presents a two-layer, human-centric production planning framework designed to optimize both operational efficiency and workforce fairness in industrial manufacturing. The first layer formulates the Order-Line allocation as a Constraint Programming (CP) problem, generating high-utilization production schedules that respect machine capacities, processing times, and due dates. The second layer models Worker-Line allocation as a Markov Decision Process (MDP), integrating human factors such as worker preference, experience, resilience, and medical constraints into the assignment process. Three solution strategies, greedy allocation, MCTS, and RL, are implemented and compared across multiple evaluation scenarios. The proposed system is validated through 16 test sessions with domain experts from the automotive industry, combining quantitative key performance indicators (KPIs) with expert ratings. Results indicate that the CP-based scheduling approach produces compact, feasible production plans with low tardiness, while the MDP-based worker allocation significantly improves fairness and preference alignment compared to baseline approaches. Domain experts rated both the Order-Line and Worker-Line components as effective and highlighted opportunities to further refine the objective function to penalize excessive earliness and improve continuity in worker assignments. Overall, the findings demonstrate that combining CP with learning-based decision-making provides a robust approach for human-centric production planning. The approach enables simultaneous optimization of throughput and workforce well-being, offering a practical foundation for fair and efficient manufacturing scheduling in industrial settings.

Figures (9)

• Figure 1: Two-layer optimization framework. The first layer uses a CP solver for Order–Line Allocation, producing a schedule of which geometries are produced on which machines and when. The second layer uses an MDP-based approach for Worker–Line Allocation, integrating the schedule with shift plans and human factors to assign workers fairly and efficiently.
• Figure 2: Decision variables introduced for a batch of geometries $g$ in the Order–Line Allocation model. Each geometry may be produced on multiple lines, with production time depending on throughput and setup times.
• Figure 3: Illustration of a CP solution. For each geometry, one alternative from the feasible set is chosen such that the resulting rectangles do not overlap on the time axis.
• Figure 4: Construction of time intervals $\delta_n$ from task and shift boundaries. White circles indicate personnel slots that must be filled in subsequent Worker–Line Allocation.
• Figure 5: Example state sequence for solving a Worker-Line Allocation problem, illustrating the evolution of allocations across successive MDP steps. Transitions $s_0 \rightarrow s_1$, $s_{15} \rightarrow s_{16}$, $s_{82} \rightarrow s_{29}$, and $s_{30} \rightarrow s_{83}$ highlight the assignment of a worker to an empty slot (white circle with red dashed border), with $s_0$ also showing the progression from $\delta_1$ to $\delta_2$.