A Combined Modelling Approach for Multi-Agent Collaborative Planning in Global Supply Chains

This paper describes the development of a combined collaborative planning model for global supply chains under environmental dynamics and uncertainties. The proposed approach is based on multi-agent modeling of global supply chains, and employs a combination of Q-learning, system dynamics, and Bayesian networks in a multi-agent architecture. It is enhanced by heuristics information from fluctuation estimations of the critical states context of supply chain networks such as process lead-time and demand variables. The heuristics are used as constraint information for aligning local agents planning with respect to system global goals, and to navigate the trade-offs between satisfying demand service levels and cost goals through the interaction and coordination between agents. Through multi-agent simulations, we compare and analyze the corresponding collaborative tactics on global supply chain planning with different products and for different environmental conditions.