Due to the flexibility, ease of deployment, and availability of various sensors, UAV-assisted mobile crowdsensing (MCS) systems have played an important role in sensing tasks. In such systems, we need to determine the allocation of UAVs to sensing tasks efficiently. However, the limited battery capacity of UAVs remains a critical constraint, significantly impacting the ability of these systems to complete sensing tasks. Furthermore, sensing tasks may have a time constraint, where tasks have defined start and end times. These factors increase the complexity of sensing task allocation in such systems. In this paper, we explore the scenario that extends UAV operational time by deploying charging stations while simultaneously considering the time constraints of the tasks. To address the task allocation problem within a time-constrained MCS environment with multiple UAVs and charging stations, we propose a novel task allocation approach, named Multi-UAV Task Allocation Algorithm with Charging Stations (MUTACS). MUTACS determines UAV trajectories to efficiently complete assigned tasks. The current task allocation decision may affect the task allocation the next time, and thus the multi-UAV task allocation problem is a sequential decision-making problem. We model it as a partially observable Markov decision process and solve it through the Multi-Agent Deep Deterministic Policy Gradient (MADDPG) algorithm. Experimental results from both synthetic and realistic datasets demonstrate that the deployment of charging stations significantly enhances the task completion capability of the MCS platform. Specifically, MUTACS achieves a <InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(29.32\%\)</EquationSource> </InlineEquation> improvement in task completion over MUTA, which does not utilize charging stations. Additionally, MUTACS consistently outperforms other algorithms that incorporate charging stations, showing an average improvement of <InlineEquation ID="IEq2"> <EquationSource Format="TEX">\(13.90\%\)</EquationSource> </InlineEquation> over DDPG, <InlineEquation ID="IEq3"> <EquationSource Format="TEX">\(10.27\%\)</EquationSource> </InlineEquation> over Greedy-P, and <InlineEquation ID="IEq4"> <EquationSource Format="TEX">\(28.88\%\)</EquationSource> </InlineEquation> over Greedy-T in terms of task completion.

A Task Allocation Algorithm in Mobile Crowdsensing Under Time Constraints

期刊论文
发表日期：

2026年3月3日

查看全文

Qijia Wang,
Bing Shi

摘要

Due to the flexibility, ease of deployment, and availability of various sensors, UAV-assisted mobile crowdsensing (MCS) systems have played an important role in sensing tasks. In such systems, we need to determine the allocation of UAVs to sensing tasks efficiently. However, the limited battery capacity of UAVs remains a critical constraint, significantly impacting the ability of these systems to complete sensing tasks. Furthermore, sensing tasks may have a time constraint, where tasks have defined start and end times. These factors increase the complexity of sensing task allocation in such systems. In this paper, we explore the scenario that extends UAV operational time by deploying charging stations while simultaneously considering the time constraints of the tasks. To address the task allocation problem within a time-constrained MCS environment with multiple UAVs and charging stations, we propose a novel task allocation approach, named Multi-UAV Task Allocation Algorithm with Charging Stations (MUTACS). MUTACS determines UAV trajectories to efficiently complete assigned tasks. The current task allocation decision may affect the task allocation the next time, and thus the multi-UAV task allocation problem is a sequential decision-making problem. We model it as a partially observable Markov decision process and solve it through the Multi-Agent Deep Deterministic Policy Gradient (MADDPG) algorithm. Experimental results from both synthetic and realistic datasets demonstrate that the deployment of charging stations significantly enhances the task completion capability of the MCS platform. Specifically, MUTACS achieves a \(29.32\%\) improvement in task completion over MUTA, which does not utilize charging stations. Additionally, MUTACS consistently outperforms other algorithms that incorporate charging stations, showing an average improvement of \(13.90\%\) over DDPG, \(10.27\%\) over Greedy-P, and \(28.88\%\) over Greedy-T in terms of task completion.