Integrated multi-channel supply chain planning with human resource and multi-tier discount considerations: An RL-GWO optimization approach

This study addresses the complex challenge of optimizing a three-level, two-channel supply chain by integrating pricing, discounting, profit allocation, ordering, and human resource management under dynamic conditions. A mixed-integer nonlinear programming (MINLP) model is proposed, incorporating price-dependent demand, channel leakage rates, multi-tier discount structures for raw material suppliers and chain stores, and human resource capacity constraints. To tackle this intricate problem, a novel Reinforcement Learning-Enhanced Grey Wolf Optimizer (RL-GWO) algorithm is developed, utilizing Q-learning to dynamically balance exploration and exploitation. Computational results demonstrate RL-GWO's superior performance, achieving near-optimal solutions with error gaps below 0.8 % compared to the BARON solver while solving large-scale instances where BARON becomes computationally infeasible. Compared to Genetic Algorithm (GA) and Particle Swarm Optimization (PSO), RL-GWO consistently delivers higher accuracy and faster computation times across all tested problem sizes. Sensitivity analyses reveal critical insights into the impacts of transportation costs, potential demand, and demand leakage on profitability, highlighting the negative effect of increased transportation costs and demand leakage on margins, while higher retailer demand significantly boosts overall profit. This research provides a robust framework and an efficient, scalable solution methodology for optimizing multi-channel supply chains with integrated human resource considerations, offering valuable strategic guidance for practitioners.