Volatile market disruptions induce abrupt changes in demand and prices, substantially complicating inventory decisions in dynamic settings. This paper develops a lost-sales multi-period inventory model that explicitly incorporates stochastic market shocks. Demand is characterized by a location-scale distribution, while market shocks are modeled as a compound Poisson process that drives the intertemporal evolution of demand parameters. Within a profit-oriented newsvendor framework, we analyze how such shocks affect ordering decisions and inventory performance under both myopic and forward-looking perspectives. To address the resulting non-stationary dynamic program, we propose a shock-aware dynamic programming approach to the optimal shock-aware policy for real-time shock hedging. We benchmark the proposed policy against several commonly used alternatives, including myopic, Gaussian variance-inflated, and shock-observed reactive policies. Numerical experiments demonstrate that our proposed approach improves expected discounted revenues by from 4% to 6% on average, and yields substantial performance gains in certain scenarios. Overall, the results underscore the importance of modeling abrupt market shocks and provide actionable insights for inventory management in highly volatile environments.
In pharmaceutical logistics, strict compliance with first-in-first-out and batch segregation requirements limits bin reuse and induces bin coupling in goods-to-person picking systems with multiple-level shuttle-based storage and retrieval. Competition for shared bins across picking stations causes frequent bin replacements, longer waiting times, and reduced picking efficiency. To address these challenges, this study develops a mixed-integer programming model for multiple picking station scheduling that explicitly characterizes bin coupling and jointly optimizes order assignment, order sequencing, and bin sequencing. The model jointly minimizes the bin replacements and parallel coupling conflicts. Two solution approaches are proposed: an exact algorithm that integrates dynamic programming with a backtracking search, and a beam search-based heuristic algorithm. Computational results show that the exact algorithm solves small-scale instances within milliseconds, whereas the heuristic algorithm provides high-quality solutions for large-scale instances within seconds. A real-world case study indicates that compared with current random or similarity-based strategies, the proposed method reduces the total cost by up to 22 percent and substantially reduces coupling conflicts. These findings provide actionable guidance for optimizing multiple picking station scheduling in pharmaceutical logistics.
This paper addresses an integrated tri-objective scheduling and internal logistics optimization problem in a U-shaped flexible flow shop under operational uncertainties, including raw material delays, setup times, and machine/mold failures. From a systems science perspective, the coupled production-logistics system is modeled as a mixed-integer nonlinear program (MINLP) that jointly minimizes: earliness-tardiness penalties, weight-dependent loading and transportation times, and congestion-driven logistics delays. A linearized reformulation is solved exactly via GAMS for small-to-medium instances. For large-scale problems, a problem-specific Simulated Annealing (SA) meta-heuristic is developed. Using real world data from an automotive rubber-metal parts supplier, the exact method optimizes instances up to 24 tasks and 6 logistics units within similar to 30 minutes (1% optimality gap). For larger instances, SA achieves average gaps below 9% with substantial runtime reductions. Uncertainty is managed via 50 independent scenarios using Sample Average Approximation. Statistical tests (paired t-test, ANOVA) confirm robustness and scalability. Benchmarking against an Improved Hybrid Genetic Algorithm (IHGA) demonstrates the competitive performance of the proposed SA. Sensitivity analyses quantify trade-offs among throughput, congestion, resource availability, and due-date slack. Key contributions include an integrated production-logistics model under multiple uncertainties, a tri-objective MINLP capturing timeliness, transport effort, and congestion, and a tailored SA for industrial-scale systems.
