An Improved Ant Colony Optimization with Adaptive Pheromone Control for Emergency Milk-Delivery Vehicle Routing: A Case Study of Durgapur

Abstract

 Timely, feasible delivery of perishable dairy products under duty-time and driver-fatigue restrictions is a practically important variant of the capacitated vehicle routing problem (CVRP). Standard Ant Colony Optimization (ACO), while effective at minimizing raw travel distance, tends to stagnate on rugged, penalty-dominated landscapes induced by non-linear fatigue and time-window penalties, frequently converging to solutions that violate operational constraints. This paper proposes an Improved Ant Colony Optimization (IACO) that combines (i) adaptive control of the pheromone evaporation rate and the pseudo-random-proportional exploitation parameter, (ii) MAX-MIN-style pheromone bounding to preserve search diversity, and (iii) a candidate-list-restricted Or-opt local search that complements classical 2-opt. The method is evaluated on a 40-node emergency milk-delivery network modelled on the Durgapur (West Bengal, India) road network, under capacity, maximum-route-time and driver-fatigue constraints. Across 15 independent seeds, IACO reduces the mean constraint-violation penalty by 22.2% and the mean total fitness (distance plus penalty) by 20.5% relative to baseline ACO (paired t-test, p=0.010 and p=0.011 respectively; Wilcoxon signed-rank, p=0.015 for both), while raw route distance is statistically indistinguishable between the two algorithms (p=0.28). The results indicate that the proposed adaptive/diversity mechanisms primarily improve constraint satisfaction rather than distance minimization, a genuine and previously undocumented trade-off for this class of problem.