Activité dans le cadre du semestre thématique sur la mobilité durable.
Cette activité est précédée par "Contextual Preference Distribution Learning for Improved Efficiency and User Experience in Ridesharing".

SVP vous inscrire pour participer à l'activité ici.

Last-mile parcel delivery is a major contributor to urban energy use and delivery-related emissions. We study a multimodal, cost-aware delivery system that combines electric vehicles (EVs), robot-assisted service, and customer self-pickup, modeled as the Two-Echelon Robot-Assisted Electric Vehicle Routing Problem with Pickup Stations and Customer Preferences (2E-REVRP-PSCP). We propose a compact Mixed-Integer Linear Programming (MILP) formulation and a two-stage matheuristic that couples a Simulated Annealing Adaptive Neighborhood Search (SA-ANS) with a two-index EV routing optimizer. Experiments on small, medium, and large benchmarks show strong scalability and solution quality, including improvements over available best-known values on large instances. Sensitivity and scenario analyses provide sustainability-oriented insights on siting, pickup-station coverage, EV range/capacity thresholds, carbon emission, and service trade-offs, supporting practical deployment in dense urban and retailer–locker networks.