Next-Gen Data Analytics and Intelligent Automation
An Intelligent Multi-Agent Architecture for Autonomous Supply Chain Decision-Making in Large-Scale Retail Systems
Authors: Suresh Kumar
Publishing Date: 13-10-2026
ISBN: 978-81-975670-6-3
Abstract
Large retail networks must reconcile local demand signals with shared inventory, transportation capacity, supplier commitments, and service objectives. A decision that improves one store can consume resources required elsewhere, while delayed information can turn individually reasonable replenishment actions into network instability. This paper proposes an intelligent multi-agent architecture that separates demand estimation, inventory planning, capacity coordination, execution authorization, and performance assurance. Intelligence is treated as the capacity to use observations, evaluate alternatives, and adapt within explicit authority boundaries; it does not require every agent to use a large language model. The architecture combines local decision autonomy with regional resource arbitration, a versioned operational state, and a deterministic execution gateway. A mathematical formulation defines inventory transitions, feasible allocations, cost, and service measures. An original synthetic experiment evaluates three interpretable policies across 320 store–product streams, four operating scenarios, and 30 paired random seeds. A further capacity sensitivity experiment tests five supply ratios. Under combined promotion and capacity disruption, adaptive proportional allocation achieves a mean normalized cost of 0.6404 artificial cost units per requested unit and a 93.397% fill rate. A greedy urgency coordinator produces a higher cost of 0.7129 and a lower fill rate of 92.603%. These findings reject the assumption that adding coordination automatically improves performance and motivate explicit policy comparison before deployment. The contribution is an auditable architecture, a bounded numerical demonstration, and a validation protocol for expanding toward large retail networks. The study does not claim a deployed system, trained multi-agent reinforcement learning model, or commercial savings. Its principal conclusion is that autonomous retail decision-making requires both effective resource allocation and independently enforced operational constraints.
Keywords
Multi-agent systems; retail supply chains; autonomous decision-making; inventory allocation; artificial intelligence; simulation; operational governance.
Cite as
Suresh Kumar, "An Intelligent Multi-Agent Architecture for Autonomous Supply Chain Decision-Making in Large-Scale Retail Systems", In: Kusum Kumari Bharti and Noor Firdoos Jahan (eds), Next-Gen Data Analytics and Intelligent Automation, SCRS, India, 2026, pp. 108-137. https://doi.org/10.56155/978-81-975670-6-3-10