In this work, we study the problem of a shipper planning its operations for delivering its orders using the transportation and storage services of contractor suppliers.
The shipper needs to contract with the suppliers ahead of time because they do not provide their service in the short term.
Moreover, capacity prices on the spot market are high and the shipper cannot run a cost-effective business by buying capacity just in time to meet the demand of orders.
Uncertainty characterizes the decisions to make for computing ahead of time the shipper's plan of operations.
At planning time, customer orders are still unknown, as well as the capacity loss of suppliers and the actual prices on the spot market for the spare capacity that the shipper may need to buy.
Thus, the shipper wants to select a set of suppliers to contract with through medium-term agreements to secure sufficient capacity to meet the demand.
Each agreement sets a fixed fee for engaging the supplier, the price to pay for the booked capacity, as well as the compensation the shipper pays to the supplier for the capacity booked, but not used, and the compensation the supplier pays to the shipper for the capacity finally not supplied.
The shipper wants to diversify the risk of supply shortage by selecting more than one supplier and minimizing the expected costs of future operations.
The time horizon of the shipper's problem is two-stage by definition, the shipper selects the suppliers in the first stage and ensures that all orders are shipped in the second one by packing all the orders into the supplied capacity and that one the shipper buys on the spot market (if needed).
The two-stage capacity planning problem of the shipper is a variant of the stochastic variable cost and size bin packing problem that minimizes the expected total cost of the shipper's operations.
For this problem, we provide a two-stage stochastic integer programming formulation and a solution framework based on state-of-the-art stochastic optimization techniques.

Service network design is inherently a stochastic problem involving decision-making in an environment with incomplete information. The aim is to tactically design a transportation plan to efficiently and profitably move origin-destination demand through a network of transportation services in a consolidation-based freight transportation context (e.g., LTL-carriers, railroads, maritime liner navigation). The applicability of the resulting plan should consistently maintain the required high-quality service standards, crucial for customer satisfaction, despite the multiple sources of uncertainty that may impact its performance. Our focus lies on travel times, i.e., the time required by a service to travel from one point to another, the fluctuation of which can have ripple effects on the feasibility and profitability of the plan, hindering efficient consolidation and jeopardizing adherence to delivery deadlines, thus negatively impacting the carrier's reputation and revenues. We will describe a two-stage stochastic programming model prescribing the selection of services in the first stage when only probability distributions for travel times are available, and estimating costs (related to delays in operations, violating due dates in customer contracts, outsourcing) of such decisions in the second stage, where service operations are defined given travel time observations. The aim of the model is minimizing operational costs. The proposed formulation is defined on a compact network that models time continuously and, thus, mitigates the computational difficulties associated to traditional time-space networks. Preliminary results regarding the complexity of the problem and the efficiency of the formulation, together with hints on policies to hedge time fluctuations will be presented.

Corridor-based logistics service providers facilitate freight exchanges between two specific regions (e.g., Southwest China and Canada) by orchestrating local operations in each region and the long-haul corridor connecting them. These logistics service providers work with multiple independent entities, including shippers (such as producers, wholesalers, and retailers) who need to transport items from one region to another, and carriers who offer services such as long-haul transportation, local transportation within each region, sorting and consolidating at terminals, and storage at warehouses.

In this study, we address the integrated, time-dependent operational planning of such a system, seeking to synchronize activities across the different parts of the system over a planning horizon. The problem focuses on three main components: route planning for local vehicles in one region, consolidating shipments on multimodal long-haul scheduled services for transfer along the corridor, and route planning for local vehicles in the other region.

To efficiently solve this complex problem, we present the Pattern-Guided Search (PGS) heuristic. This heuristic maintains a diverse, high-quality set of elite solutions and guides the search using patterns extracted from this elite set. PGS utilizes a two-layer local search, pattern-guided perturbation, and diversification mechanisms to achieve high performance. PGS significantly outperforms general solvers (such as Gurobi) on medium and large-sized problem instances. We conducted extensive experiments and derived practical insights for decision-makers in the operational planning of corridor-based logistics systems.

Modern transportation and Supply Chain systems require a holistic view of the sustainability of their operations, considering different aspects (operations, economies of scale, social and environmental impacts) while finding a trade-off between maximization of the revenues and high quality of service for customers.
One of the principal modeling frameworks in this context turned out to be the Service Network Design. Some aspects of it, however, may limit its applicability. One of those is the absence of explicit packing aspects related to the services to be operated on the network. In this talk, we discuss new variants of service network design, namely the Service Network Design with Bin Packing Considerations, that incorporate the families of packing problems already in use in capacity planning. We will present a unified problem setting addressing simultaneously decisions on the selection of scheduled services (on a general time-space network), the selection of the type and number of capacity units to load and move the origin-destination demands, the assignment of demand flows to the loading units, and the construction of the demand itineraries within the selected service network.