Task Assignment Scheme Designed for Online Urban Sensing Based on Sparse Mobile Crowdsensing

• Published in: IEEE internet of things journal Vol. 12; no. 11; pp. 17791 - 17806
• Main Authors: Zeng, Hongjian, Xiong, Yonghua, She, Jinhua, Yu, Anjun
• Format: Journal Article
• Language: English
• Published: Piscataway IEEE 01.06.2025; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Algorithm optimization; Automation; Constraints; Costs; Crowdsensing; Cycle time; Data collection; Data integrity; Effectiveness; Evolutionary algorithms; Evolutionary computation; Genetic algorithms; Inference; Inference algorithms; Internet of Things; Mobile computing; Optimization; Sensors; sparse mobile crowdsensing (SMCS); subarea evaluation; task assignment; Time factors
• ISSN: 2327-4662; 2327-4662
• DOI: 10.1109/JIOT.2025.3540501

Sparse mobile crowdsensing (SMCS) achieves urban-scale environmental sensing by assigning tasks to workers in specific subareas and inferring global data from the collected information. However, the effectiveness of SMCS is often limited because many studies overlook workers' mobility and data collection time during subarea selection, as well as the time constraints of the sensing cycle in task assignment. This may affect the task completion timeliness and data quality. To address these issues, we develop a subarea evaluation method based on deep reinforcement learning, considering both the temporal effectiveness of sensing tasks and the importance of subarea selection for data inference. Using the subarea evaluation values derived from this method, we establish an online urban sensing task assignment model which is subject to constraints of sensing cycle time and cost budget. This model aims to find the task assignment result that minimizes data inference error by maximizing the comprehensive utility value. Considering the characteristics of the task assignment model, we propose an evolutionary algorithm named OTA-EA, which is based on an improved genetic algorithm. Its enhanced evolutionary operators can avoid generating infeasible solutions while maintaining robust search and optimization performance. Lastly, we conduct experimental evaluations of these methods on the real-world datasets. The results demonstrate that our subarea evaluation method can significantly reduce the data inference error, and our evolutionary task assignment algorithm can achieve better task assignment results than the baseline algorithms.