WB11 - Chaîne d'approvisionnement / Supply Chain

Determining optimal capacities for refurbishing, repairing, and recycling electric vehicle batteries is a complex planning problem. Economic incentives can distort allocation across processing streams, while deterministic approaches fail to capture uncertainty in supply, battery quality, and service-level requirements. This study develops a stochastic optimization model for a dynamic circular-economic context.
Keywords: Electric vehicle batteries, capacity planning, stochastic programming

Omnichannel retailers must jointly manage product assortment, inventory, and customer substitution across regions and channels under demand uncertainty. We develop a two-stage stochastic mixed-integer linear program that simultaneously optimizes SKU activation, multi-period inventory replenishment, and substitution flows as recourse decisions. Demand is modeled endogenously through an iterative procedure embedding the Generalized Attraction Model, capturing assortment-dependent demand redistribution while preserving MILP tractability. A rolling horizon framework enables period-by-period re-optimization as demand information accumulates. The model incorporates cross-channel demand shifting and conditional value-at-risk for risk-averse planning. To evaluate robustness, we propose a structured stress-testing methodology covering demand surges, facility outages, substitution drops, and channel preference shifts. We will present computational experiments on synthetic datasets to evaluate solution quality, scalability of the decomposition approach, and the behavior of adaptive re-optimization under disruption scenarios.

We examine a dynamic vertical supply chain in which upstream and downstream firms coordinate through a dynamically adjusted cost-plus contract. Demand is always satisfied via flexible production, so outsourcing emerges endogenously as the residual between realized demand and installed capacity. We compare cooperative, open-loop Nash, and closed-loop Nash equilibria to investigate how strategic interaction shapes pricing, capacity investment, and outsourcing. Our core insight is that outsourcing arises primarily from information-induced coordination failure rather than cost advantages, flexibility motives, or pricing distortions alone. Under cooperation or open-loop commitment, firms align intertemporal decisions, sustaining high in-house product availability and minimal outsourcing, even under demand uncertainty. In contrast, closed-loop (feedback) competition induces continuous strategic adjustments that erode investment incentives, resulting in persistent undercapacity and higher outsourcing. Multiplicative demand uncertainty amplifies this effect in closed-loop equilibria, strictly increasing outsourcing and reducing in-house product availability, while cooperative and open-loop regimes remain robust, with in-house availability virtually unaffected. Asymmetric strategy profiles (one firm open-loop, the other closed-loop) further exacerbate outsourcing imbalances. In the limit of extreme uncertainty, all equilibria converge to the same myopic outcome, neutralizing strategic distinctions. These findings reinterpret outsourcing as a structural byproduct of intertemporal non-commitment and information-induced coordination failure in decentralized supply chains, highlighting the critical role of commitment mechanisms in preserving efficient capacity alignment.

In this study, we address the multi-period Production Routing Problem (PRP) in a two-echelon supply chain system with simultaneous pickup and delivery. The problem consists of multiple competing supply chains engaging in potential horizontal collaboration. Each supply chain includes multiple production plants, distribution centers (DCs), and retailers, and involves two transportation layers: from plants to DCs and from DCs to retailers.
Each transportation echelon is modeled as a Vehicle Routing Problem with Simultaneous Pickup and Delivery (VRPSPD) using a heterogeneous vehicle fleet, in which vehicles deliver products to downstream facilities while collecting returned packages. The collected packages are transported back to the DCs and subsequently returned to the production plants.
The main objective of this research is to evaluate the impact of horizontal collaboration among competing supply chains compared to a fully non-collaborative system. Collaboration is examined at three different levels of the supply chain: (1) Production level, through shared production capacities. (2) Distribution centers level, through shared inventories for both products and packages. (3) Transportation level, through shared vehicle fleets.
To analyze these configurations, we develop mixed-integer linear programming (MILP) models for the non-collaborative case and for each collaboration level. The models aim to minimize the total operational cost over the planning horizon, including production, inventory holding, and transportation costs. The results analysis enables us to identify the collaboration strategy that provides better operational benefits.