A new inventory routing problem with route and schedule unpredictability

The inventory routing problem (IRP) in a retail supply chain setting allows for the simultaneous optimisation of delivery schedules, vehicle routes, and delivery quantities. The IRP relies on the adoption of a vendor-managed inventory strategy which has the potential to reduce transportation, inventory, and stock-out costs in a supply chain. In this paper, we introduce a mathematical model for a new IRP variant, the heterogeneous fixed fleet IRP with time-windows (HeFIRPTW) with route and schedule unpredictability, in the form of a bi-objective mixed-integer linear programming problem. This model simultaneously incorporates route and schedule unpredictability aimed at mitigating inherent safety and security threats experienced during the transportation of valuable goods. Delivery routes and schedules are generated that minimise the operational costs incurred whilst also ensuring that route segments are not traversed too regularly and that customers are not visited during overlapping daily time intervals. The feasibility of adopting an exact ϵ -constrained model solution solution method is investigated empirically by solving small, adapted benchmark instances of the problem. An investigation into the model solution complexity for varying problem sizes reveals that unpredictability, particularly with tightened constraints, increases the computational time. The complexity implications of multiple vehicles and the imposition of time-windows are also examined. The results highlight the computational demands of the proposed model, demonstrating a clear need for a faster, perhaps approximate, solution approach capable of generating high-quality solutions for realistic problem instances within reasonable time-frames. • An inventory routing problem with route and schedule unpredictability is proposed. • A novel bi-objective, mixed-integer linear programming model is formulated. • The model is validated in the context of a benchmark data set. • An iterative ϵ -constraint method is adopted to solve problem instances. • The time complexity of adopting an exact solution approach is studied empirically.