Modern logistics networks face escalating complexity driven by exponential e-commerce growth, consumer demand for rapid last-mile delivery, and the imperative for sustainable operations. Traditional centralized optimization approaches struggle to scale effectively across dynamic, large-scale distribution networks while maintaining real-time adaptability. This dissertation addresses these challenges by designing, implementing, and evaluating a novel Hierarchical Multi-Agent Swarm (HMAS) framework that integrates Particle Swarm Optimization (PSO) at the strategic planning layer, Ant Colony Optimization (ACO) at the tactical routing layer, and reactive autonomous agents at the operational coordination layer for autonomous logistics optimization. Employing a post-positivist, simulation-based experimental methodology grounded in Design Science Research, the study evaluated the HMAS framework across four publicly available datasets: CVRPLIB benchmark instances, NYC Taxi and Limousine Commission trip records, OpenStreetMap via the Open Source Routing Machine, and the NCO-VRP dataset spanning 100 global cities. Results demonstrated that the HMAS framework achieved a mean optimality gap of 2.65% on CVRPLIB benchmarks, significantly outperforming standard ACO, standard PSO, Adaptive Large Neighborhood Search, greedy heuristics, and Google OR-Tools baselines. The Friedman test confirmed significant omnibus differences (χ²(5) = 121.54, p < .001), and Wilcoxon signed-rank tests with Holm correction revealed large effect sizes favoring HMAS across all pairwise comparisons (d = −1.02 to −3.29). The hierarchical architecture maintained sub-exponential computation time growth across problem sizes from 25 to 500 nodes, achieved 86.6% on-time delivery rates under dynamic rush-hour demand derived from NYC taxi patterns, and generalized consistently across 10 heterogeneous city topologies with a mean cross-city optimality gap of 2.70%. All four hypotheses were supported, demonstrating that hierarchical swarm coordination offers a scalable, adaptive, and generalizable paradigm for autonomous logistics optimization with implications for both swarm intelligence theory and industry practice.
Laszlo Pokorny (Fri,) studied this question.