A Markov Decision Process Approach to Dynamic Re-Routing and Inventory Pre-positioning under Stochastic Supply Chain Disruption

Abstract

 Anomalous disturbances including transport link failures, facility closures and abrupt capacity losses significantly impair the performance of supply chains. Static network structures and constant safety-stock policies are not adaptive to the development of unexpected disruption patterns. To the best of our knowledge, this paper is the first to address shipment re-routing and inventory pre-positioning simultaneously in an integrated way for a multi-echelon supply chain under repeated stochastic disruptions. The network is represented as a Markov Decision Process with state comprised of node-level inventories, backorders and link status and actions determining shipment amounts on open routes and pre-positioning at strategic nodes. The aim is to minimise long-term expected total cost, subject to service-level constraints. Owing to high state–action dimensionality, DRIP-MDP is solved by applying simulation-based approximate dynamic programming with value function approximation. The performance is measured based on five metrics: the expected total cost (×10⁵), the service level (%), unsatisfied demand (the units/period), average inventory (units) and a resilience index. Comparative studies involving four established policies (SS-FR, RTSP, IOP, MRH) demonstrate that DRIP-MDP attains the lowest cost (0.78 compared to 1.00, 0.92, 0.95, and 0.89), the highest service level (97.8% versus 93.0–95.2%), and the least unmet demand (55 units/period versus 120, 100, 95, and 90).  The average inventory under DRIP-MDP is 1210 units, which is comparable to the baseline range of 1180–1400 units, while the resilience index has improved to 0.91, in contrast to the 0.78–0.85 range of existing approaches.  The results indicate that DRIP-MDP offers enhanced cost efficiency, service performance, and resilience while avoiding unnecessary inventory expansion, hence presenting a viable solution for disruption-aware supply chain operations.