Evolutionary Multi-Task Allocation For Mobile Crowdsensing With Limited Resource

Multi-task allocation, as a crucial issue on mobile crowdsensing (MCS), is to find the optimal participants-task pair under limited resource. The previous assignment methods normally maximize the benefit of participants or task requester, giving no consideration to the comprehensive interests of all parties in MCS. Besides, the sensing ability of a mobile user is commonly evaluated by the detection quality or the probability of covering the target area, causing the unqualified measurement with low spatial coverage degree or bad accuracy. To overcome these drawbacks, a comprehensive multi-task allocation model for MCS is formulated, in which maximizing the overall sensing quality of all tasks is the objective and three constraints come from the total budget shared by all tasks, the sensing quality of each task and the workload of each mobile user. An aggregative indicator that integrates the spatial coverage degree with the sensing accuracy is presented to estimate the overall quality of a mobile user. Following that, a large-scale evolutionary algorithm with problem-specific repair strategy and the new genetic operators is presented as a problem-solver. The former generates the superior feasible solutions for the overloaded participants and the unqualified tasks, while the latter introduces the new promising users to explore the high dimension decision space, with the purpose of speeding up the convergence and avoiding falling into the local optimum. The experimental results based on 30 test instances indicate that the proposed evolutionary multi-task allocation method can generate the superior feasible assignments as more as possible, which outperform the state-of-the-art algorithms.