Energy-efficient Q-learning-based path planning for UAV-Aided data collection in agricultural WSNs

Abstract

Recently, unmanned aerial vehicles (UAVs) have emerged as a promising solution for efficient data collection in agricultural Wireless Sensor Networks (WSNs). However, rely on a single UAV faces significant challenges due to vast coverage areas, energy constraints, and connectivity limitations, leading to scalability and energy inefficiency issues. This paper proposes a multi-UAV-based data collection approach, leveraging Q-learning (QL)-based UAV path planning to optimize trajectory selection dynamically while addressing the challenges of minimizing energy consumption and travel distance. We model the UAV path planning as a variant of the Traveling Salesman Problem (TSP), where UAVs start and end their routes at a ground base station (GBS) after visiting designated data collection points (CPs). Unlike traditional heuristic-based methods, the QL model adapts in real-time to environmental and operational conditions, ensuring fault tolerance and operational robustness. Through extensive simulations using NS3, our results demonstrate that deploying multiple UAVs instead of a single UAV improves mission completion time by approximately 67% and increases residual energy levels by more than 80% when managing up to 500 sensor nodes with four UAVs. Furthermore, the proposed QL-based path planning significantly reduces total travel distance and energy consumption, enabling balanced workload distribution among UAVs and ensuring higher data collection reliability in large-scale agricultural WSNs. These findings confirm the effectiveness of a multi-UAV, QL-based approach for scalable, fault-tolerant, and energy-efficient UAV-aided data collection systems.