Online food delivery and supply chain management by optimization of innovative models

Abstract The study develops an empirically validated simulation framework to optimize urban food delivery through vehicle differentiation, multi-order management, and shortest-path routing. Using an empirical dataset from a medium-sized European city, three models are compared: single-order, greedy multi-order, and optimized multi-order delivery. Routing combines Dijkstra and Traveling Salesman Problem heuristics, with travel times predicted by a k-Nearest Neighbors model. The simulation closely reproduces observed operational patterns and shows that multiple-order delivery reduces average delivery time by 15 %–20 %, while the optimized model achieves an additional 8 % improvement. Outcomes indicate higher operational efficiency, better vehicle utilization, and reduced idle time. Overall, the approach supports both economic and environmental sustainability in last-mile logistics, offering a realistic benchmark for algorithmic coordination in heterogeneous delivery fleets offering better customer satisfaction.