|
Unmanned Aerial Vehicles (UAVs) serve as mobile sensors for real-time collection of spatio-temporal passenger flow data within regions, advancing the development of regional intelligent transportation systems. By hovering and capturing video, UAVs continuously monitor new passenger arrivals at points of interest (POIs), transmitting this data to a remote control center. POIs are locations with monitoring demands, referred to as demand points. A cost-effective solution requires the dynamic deployment of limited UAV resources to maximize monitoring performance by collecting the most comprehensive monitoring data possible. Given the dynamic and stochastic nature of passenger flows, a multi-UAV situational awareness and decision-making system for dynamic scheduling is proposed. Situational awareness involves obtaining overall status and using obtained monitoring data for estimation of approximate passenger flow and calculation of real-time monitoring demands. Decision-making entails routing UAVs, modeled as a Dynamic and Stochastic UAV Task Assignment Problem (DSUAVTAP), employing a greedy algorithm for online problem-solving. Using randomly generated datasets, a genetic algorithm is employed to calibrate five unknown parameters in the monitoring demand function. Experiments with 20 different passenger flow configurations compare dynamic UAV scheduling to static monitoring loops. Results show that dynamic scheduling outperforms static scheduling by an average of 9.5% in monitoring performance. Therefore, dynamic UAV scheduling proves superior for monitoring regional passenger flows with the fluctuating and uncertain characteristics. This work lays a foundational framework for future advancements in intelligent UAV scheduling techniques and optimal allocation of UAV resources in regional monitoring.
|
