AI-Based Dynamic Waste Collection Route Optimization for Smart Cities

Abstract

The waste management of a Smart City must follow intelligent routing systems that multi-objective, unlike purely distance minimization, and incorporates urgency signal from IoT monitored bins, and UAV derived allocation of waste prioritization. This paper proposes the PHGA-ACO framework that focuses on dynamic waste collection route optimization. Subsequently, the proposed framework efficiently integrates the smart bin fill-level data (Paper 1). In addition, the work integrates the WPI-HES outputs (Paper 2). The model works with a Capacitated Vehicle Routing Problem (CVRP) on a real OpenStreetMap road network of Nashik Municipal Corporation, India, having 84 active collection nodes with total demand 24,498.9 kg. A systematic comparison of five baseline algorithms First Come First Serve (FCFS), Nearest Neighbor (NN), Dijkstra-greedy, Ant Colony Optimization (ACO), and Genetic Algorithm (GA) with PHGA-ACO under same operating conditions. S(m) is a composite metric that combines the normalized route distance and the Urgency Position Index (UPI) that measures the joint efficiency–priority-compliance performance of each method. PHGA-ACO obtains the best overall score (S=0.2556, the best among all policies) as it cuts down total collection distance by 35.6% compared to that of FCFS. Compared to ACO, it improves the priority servicing index by 13.4% at the expense of a mean distance increase of 68.7% compared to ACO. The improvement happens specifically due to the urgency penalty component, as seen from the ablation analysis. Sensitivity analysis indicates optimal urgency weight range λ ∈ [54.3, 162.9]. Statistical analysis of 30 independent runs with the Wilcoxon and Friedman test indicates that all inter-algorithm differences are significant (p < 0.001).