Autonomous Delivery Vehicles (ADVs) and Shared Delivery Locations (SDLs) provide alternative service options for urban last-mile logistics, but their effectiveness depends on customer acceptance and network design. To capture these features, we formulate the Time-Dependent Heterogeneous Vehicle Routing Problem with Delivery Options (TD-HVRPDO), which jointly models autonomous and manned fleets, home and SDL delivery modes, customer-specific acceptance constraints, and time-dependent travel times under urban congestion. The model jointly determines delivery mode, vehicle type, and routing decisions while accounting for the consolidation opportunities provided by shared delivery locations. We develop a hybrid matheuristic integrating Column Generation, Large Neighborhood Search, and Route-based Search. Computational experiments show that the proposed method produces high-quality solutions more efficiently than a commercial solver and adapted benchmark methods. Managerial analyses show that both ADVs and SDLs can reduce the costs, with SDLs yielding larger savings through delivery consolidation. However, the extent of these consolidation benefits depends strongly on SDL placement. Moreover, joint deployment can provide further savings while reducing reliance on manned vehicles. These findings therefore suggest that operators should coordinate investment in additional ADVs with SDL network design and efforts to broaden customer acceptance.
Modernisation of railway infrastructure can lead to significant improvements in the efficiency of train services. However, the deployment of the modernisation requires huge financial investments and is carried out over decades. Consequently, infrastructure upgrade decisions must be strategically planned over time. In this paper, we introduce the Railway Infrastructure Upgrade Deployment Problem (RIUDP) in which we optimize phased deployment of infrastructure upgrades in a railway network so that societal benefits, measured by the reduction in the travel times of all passenger trains, are maximized. This requires to compute a timetable for each potential deployment plan of the upgrades. To compute this timetable, we assume that the structure of the timetable is given.We formulate the RIUDP as a Mixed Integer Linear Program and solve it to optimality. We compare optimised solutions with solutions that mimic rule of thumb decisions based solely on the theoretical cost effectiveness of infrastructure upgrades. A detailed case study on the roll-out of ERTMS on the main Dutch railway network reveals that severe gains can be achieved using optimized timetables instead of the rule of thumb decisions. Finally, we study the impact of specific practical considerations such as regional equity and connectivity constraints on the RIUDP.
This study investigates the operational optimization of an on-demand, multi-modal relay-based delivery system comprising dispersed unmanned hub stations and fleets of drones and unmanned vehicles. We formulate a plan-based model to coordinate and synchronize heterogeneous delivery modes, ensuring timely and cost-effective fulfillment of customer orders. The proposed model determines which orders to accept in each arriving batch, allocates drones and vehicles to execute these orders, and schedules their movements to guarantee on-time deliveries. A nested exact algorithm, integrating logic-based Benders decomposition (LBBD), column-and-row generation (CRG), and branch-and-price (BP), efficiently solves typical instances in approximately 70 seconds. The algorithmic novelty lies in that a CRG & BP based exact algorithm is embedded in an LBBD to solve the LBBD’s subproblem. Computational experiments, based on realistic scenarios, demonstrate the effectiveness of our proposed methodology and provide several managerial insights. Factors such as average delivery distance, time-window flexibility, and fleet composition are shown to significantly impact system profitability and service quality. The results offer practical guidance on batch sizing, fleet configuration, and operational conditions under which multi-modal relay-based delivery systems outperform traditional courier-based models.
Effective planning of capital-intensive fleets such as tugboats requires coordinating long-term strategic decisions (e.g., fleet renewal), medium-term tactical decisions (e.g., chartering), and daily operational decisions (e.g., service-order acceptance and fleet deployment) under uncertain market conditions. Existing fleet-planning methods commonly represent only part of this strategic-tactical-operational hierarchy or simplify the operational consequences of uncertainty, potentially leading to economically suboptimal or operationally infeasible fleet decisions. To address these gaps, this paper develops a holistic methodology for dynamic, hierarchical, and stochastic fleet planning under uncertainty in service demand, costs, and revenues. First, we formulate a hierarchical Markov decision process (HMDP) in which strategic renewal decisions govern interperiod fleet evolution, while tactical chartering and operational order-acceptance and deployment decisions are optimized as conditional within-period recourse. Second, we develop a hierarchy-aware approximate dynamic programming (ADP) approach that combines rolling scenario-based approximation of the continuation value with a property- and data-informed decision-refinement mechanism for screening inferior strategic candidates. Third, we instantiate the methodology as a realistic tugboat fleet planning problem calibrated using industrial data from PSA Marine. In the baseline experiments, HMDP-ADP (H=10) achieves average expected-profit improvements of 3.67%–11.43% and maximum observed improvements of 6.39%–11.76% relative to the four benchmark methods, while remaining computationally tractable for practical-scale instances. Extensive sensitivity analyses examine the effects of strategic review opportunities, charter-market depth, service demand, fleet-base layout, deployment-modeling resolution, and forecast accuracy. Although developed for tugboat fleets, the framework is applicable to other capital-intensive maritime fleets with long-lived assets, active chartering markets, uncertain demand, and coupled operational constraints.
Disasters pose persistent threats to human life and infrastructure, making evacuation planning vital for disaster risk reduction. The effectiveness of such efforts depends on understanding and accurately modeling human behavior. This study reviews 479 peer-reviewed articles on large-scale evacuation planning and management, classifying them along two dimensions: behavioral factors influencing key authority-issued decisions and modeling approaches incorporating these dynamics within optimization, simulation, and simulation–optimization frameworks. While demographic and route-choice behaviors are extensively studied, mode selection and road-segment designation remain largely unexplored. Optimization models commonly assume full- or non-compliance, with limited research that considers partial-compliance, leading to more realistic solutions. On the other hand, simulation-based studies capture dynamic and uncertain conditions more effectively but are dominated by rule-based behavioral proxies, particularly for departure timing and route choice, and typically limit dynamic feedback to route choice alone. Simulation–optimization approaches, while increasingly adopted, remain largely reliant on parameter-based behavioral representations rather than empirically calibrated, behavior-oriented coupling. Emphasizing the importance of embedding behavioral heterogeneity in model structures, this review develops a taxonomy, highlights research gaps, and outlines future directions for behaviorally realistic evacuation models that enhance the effectiveness and practicality of evacuation strategies in real-world crises.
