Modular Freight Routing in Physical Internet-Enabled Macro and Micro Logistics Systems

The rapid expansion of global and e-commerce flows demands a fundamental redesign of freight logistics toward systems that are simultaneously efficient, resilient and sustainable. This thesis brings together two complementary research streams within the Physical Internet (PI) paradigm to deliver a unified framework that addresses freight routing at both the macro (intermodal) and micro (urban last-mile) scales. At the macro scale, the thesis formulates a centralized intermodal routing model for PI-containers that allows containers to be disaggregated into modules and routed through a network of PI-hubs and terminals. Performance is evaluated with four Key Performance Indicators, which are truck utilisation, total delivery time, transportation cost, and inter-module delivery gap, and the model is used to study system robustness under uncertainty via a variance-based Global Sensitivity Analysis (GSA). The robustness study identifies the most influential PI parameters (notably PI-hub processing times and moduleization levels) that affect operational KPIs and hence informs design choices for resilient synchromodal networks. At the micro scale, the thesis proposes a PI-inspired two-echelon van–robot delivery model for last-mile operations in dense urban areas. Customer orders are modularized into standard parcel modules; electric vans act as mobile micro-hubs that transport and deploy capacitated sidewalk autonomous delivery robots (SADRs) at drop-off points. The problem is formulated as a mixed-integer program that jointly decides van routes, robot assignments, and module allocation, with multi-objective goals to minimize fleet size, completion time of last module, total energy consumption, and intra-parcel delivery gap. To solve realistically sized instances, a novel matheuristic called: Multi-Stage Greedy-Search Decomposition (MSGSD), is developed and benchmarked against exact Branch-and-Cut and another metaheuristic baselines (biased Simulated Annealing algorithm), demonstrating superior scalability and high solution quality. By integrating macro-scale robustness insights with micro-scale operational designs, the thesis provides a cross-scale methodology for PI deployment: (i) analytical tools to identify network vulnerabilities and key design parameters at the intermodal level; and (ii) an operationally feasible, modular last-mile solution that improves service, lowers energy use, and enhances asset utilization in cities. The combined contributions offer both theoretical advances (models, solution methods, and sensitivity analyses) and practical guidance for planners and operators seeking to transition toward a modular, shared, and resilient logistics ecosystem.