A Dynamic Unmanned Aerial Vehicle Routing Framework for Urban Traffic Monitoring

Abstract

Unmanned Aerial Vehicles (UAVs) have great potential in urban traffic monitoring due to their rapid speed, cost-effectiveness, and extensive field-of-view, while being unconstrained by traffic congestion. However, their limited flight duration presents critical challenges in sustainable recharging strategies and efficient route planning in long-term monitoring tasks. Additionally, existing approaches for long-term monitoring often neglect the evolving nature of urban traffic networks. In this study, we introduce a novel dynamic UAV routing framework for long-term, network-wide urban traffic monitoring, leveraging existing ground vehicles as mobile charging stations without disrupting their operations. To address the complexity of long-term monitoring scenarios involving multiple flights, we decompose the problem into manageable single-flight tasks, in which each flight is modeled as a Team Arc Orienteering Problem with Decreasing Profits with the objective to collectively maximize the spatiotemporal network coverage. Between flights, we adaptively update the edge weights to incorporate real-time traffic changes and revisit intervals. We validate our framework through extensive microscopic simulations in a modified Sioux Falls network under various scenarios. Comparative results demonstrate that our model outperforms three baseline approaches, especially when historical information is incomplete or absent. Moreover, we show that our monitoring framework can capture network-wide traffic trends and construct accurate Macroscopic Fundamental Diagrams (MFDs). These findings demonstrate the effectiveness of the proposed dynamic UAV routing framework, underscoring its suitability for efficient and reliable long-term traffic monitoring. Our approach's adaptability and high accuracy in capturing the MFD highlight its potential in network-wide traffic control and management applications.