Randomization-Embedded Greed: Pursuing More Platform Profits in Mobile Crowd Sensing

The recent proliferation of sensor-equipped mobile phones, together with the intrinsic mobility of their users, has enabled mobile crowdsensing (MCS) to spring up. In a typical MCS system, the MCS platform recruits mobile phone users (workers) to perform the sensing tasks published by requesters. The requesters are interested in some urban events and are willing to pay for the sensing data returned by the workers. A lot of incentive mechanisms have been presented in the literature, in order to stimulate workers to participate into MCS campaigns or to maximize the social welfare. However, the community has not yet paid much attention to optimizing the platform profit, which is highly valued by the profit-making MCS organizers. In this paper, we consider a realistic MCS scenario that depends on probabilistic collaboration among workers and has constrained capacity of platform; and we focus on the platform profit maximization (PPM) in this MCS scenario. The PPM problem is NP-hard, and the key challenge stems mainly from the non-monotonicity caused by the probabilistic collaboration and the quality-based payment of the requester. First, we propose two polynomial-time approximation algorithms for the PPM problem: MaxG and RandG. Our main effort is dedicated to design RandG, which involves a randomization policy of selecting workers in its greedy framework, aimed at pursuing chances of skipping over incompetent local optima. We also prove that RandG can achieve a constant approximation ratio in expectation. Second, we present algorithm RandCom for general PPM problems, which combines MaxG and RandG to struggle to make as high platform profit as possible. Finally, we conduct extensive simulation to evaluate our designs in terms of platform profit.