Cluster based Online Task Assignment for Mobile Crowdsensing

Mobile crowdsensing has become a promising sensing paradigm with the popularization of mobile devices. In this paper, we focus on an opportunistic mobile crowdsensing scenario where there are multiple task requesters and users, who move in an opportunistic way in the target environment. When a task requester encounters a user, he can assign some of his held tasks to the user and receive corresponding task results when they re-encounter sometime later. In this paper, we study how to minimize the largest makespan of all requesters for the task result collections. To address this issue, we propose a cluster based largest makespan sensitive online task assignment (CLOTA) algorithm. C-LOTA first performs two-phase clustering which clusters the users into different clusters, one for each task requester, based on their relativeness to the task requesters and also the task workloads at different requesters. C-LOTA then iteratively performs greedy intra-cluster task assignment such that largest task is firstly assigned and the first idle user always takes the task, until all tasks are assigned. We present the detailed algorithm design of C-LOTA. We deduce its computation complexity. Simulation results show that C-LOTA can achieve much better performance compared with existing work.