Efficient Waste Collection Routing Using F-CVRP and Dynamic Parameter Optimization via Q-Learning

Efficient waste collection is a significant challenge for urban populations and industrial sectors. This paper models the routing of waste collection using the Family Capacitated Vehicle Routing Problem (F-CVRP), considering different types of waste as families, and refers to this as the Waste Collection Problem (WCP). To solve the WCP, we introduce Q-learning integrated Genetic Algorithm (GA) and Ant Colony Optimization (ACO), denoted as QGA and QACO, respectively. The QGA and QACO models dynamically adjust their parameters, enhancing the balance between exploration and exploitation, thereby improving solution quality and convergence speed. We tested our models on 24 scenarios with varying numbers of vehicles and required visitation points. The Q-learning enhanced Ant Colony Optimization (QACO) algorithm excelled, ranking best in 16 out of 24 cases (67%). Our results demonstrate that QGA consistently outperforms classical GA and QACO surpasses classical ACO, with QACO emerging as the most effective. This highlights capability of QACO to optimize waste collection routes and enhance waste management systems.