Multi-Task Oriented Participant Recruitment for Vehicular Crowdsensing

As vehicles are equipped with more sensors, there is a growth of potential for vehicles to contribute to urban crowdsensing. However, participant recruitment, the process that decides which vehicles are appropriate for specific sensing tasks with a budget constraint, is critical, especially in multi-task scenarios. To address on this problem, we first formulate the vehicle’s route and urban road network model to convert this problem to a combinatorial optimization problem. The time complexity of the optimal solution is factorial. Therefore we proposed two algorithms: Naive recruitment and Greedy recruitment to find a suboptimal solution for single task scenario. Furthermore, considering the sensing overlaps in time and regions between multiple tasks, ST-Merge is presented to merge the common sensing requirements in time and space dimensions to alleviate sensing burden. Finally, we evaluate the performance of all algorithms using real road and trajectory data in Beijing, China. The experimental results show that the combination scheme of ST-Merge and Greedy recruitment can effectively improve the sensing effect.