Vehicle Routing and Fleet Dispatch Optimization for Last-Mile Urban Distribution

Urban last-mile distribution operates within highly volatile environments characterized by stochastic
traffic congestion, rigid customer time windows, and real-time order disruptions. Traditional static
vehicle routing formulations fail to dynamically adapt to these fluctuating conditions, leading to
excessive operational costs, missed service level agreements (SLAs), and elevated fleet carbon
footprints. This paper presents an exhaustive dynamic optimization framework for last-mile logistics:
The Tabu-Guided Adaptive Large Neighborhood Search with Rollout-Based Real-Time Dispatch (T
ALNS-RRD). Building upon foundational principles in fleet dispatch literature alongside classical
routing paradigms, our architecture integrates dynamic congestion penalties with multi-layered
memory structures and predictive rollout simulations. Empirical testing across complex urban road
networks demonstrates a 24.3% reduction in total operational costs, an increase in on-time delivery
rates from 68.1% to 92.8%, and a 54.4% drop in congestion exposure compared to conventional routing
baselines.