Intelligent Route Planning for Waste Collection in Smart Cities via Reinforcement Learning

Abstract

Introduction: The increasing complexity of urban infrastructure poses challenges for efficient and sustainable municipal waste management. This paper presents a framework based on reinforcement learning (RL) for real-time route planning in smart city waste collection systems. By integrating Iot-enabled smart bins with environmental data such as bin fill levels and real-time traffic, the approach frames the routing task as a Markov Decision Process (MDP). It employs Q-learning and Deep Q-Networks (DQN) to optimise navigation policies. A simulation based on Hanoi, Vietnam, assesses the method's adaptability and efficiency under varying bin distributions and traffic conditions. Results demonstrate that DQN outperforms traditional Q-learning regarding route stability, bin coverage, and learning convergence, particularly in complex urban environments. The system minimises route length and travel time while prioritising the collection of full bins, highlighting the potential of deep reinforcement learning for scalable and sustainable waste logistics in future smart citie.