Truth based three-tier Combinatorial Multi-Armed Bandit ecosystems for mobile crowdsensing
Yingqi Peng,Wei Liu,Anfeng Liu,Tian Wang,Houbing Song,Shaobo Zhang
DOI: https://doi.org/10.1016/j.eswa.2023.121119
IF: 8.5
2023-09-03
Expert Systems with Applications
Abstract:Many Multi-Armed Bandit (MAB) based workers selection schemes have been proposed to select high-quality workers to enhance the quality of tasks. However, in Mobile Crowd Sensing (MCS), a complex mutual effect exists among task requestors, the MCS platforms, and workers. Only considering the interaction of two sides doesn't make MCS a balanced ecosystem. Therefore, it is urgent to establish a tripartite mutual incentive mechanism to make the MCS system a balanced ecosystem. In this paper, a truth based Three-tier Combinatorial Multi-Armed Bandit (TCMAB) incentive mechanism is proposed for selecting each other to maximize their revenues in MCS. In TCMAB, there exists a three-tier and two-way MAB-based incentive scheme. For the mutual interaction between the platform and the worker, a truth-based CMAB scheme is established for the platform to select high-quality workers, and also a CMAB scheme is proposed for workers to select a "good" platform to report data in order to maximize their revenue. Besides, for the mutual interaction between the task requestor and the platform, the platforms adopt the CMAB-based scheme to select a high-payment task requestor that gives high payment. And a platform selection scheme base on CMAB is also established for task requestors to select the platforms which have lower fees and higher quality. What's more, we don't adopt the assumption that platforms get the data quality as soon as they get data, but propose a data quality acquisition scheme based on the truth data discovery and cooperation frequency, which is the base to instruct the three-tier interaction, thus establish a kind of truth-based MCS interaction ecosystems. Simulation results show that the proposed TCMAB provides an effective solution for the problem of information elicitation without verification (IEWV) in MCS, and can improve the utilities, data quality, and applications quality for MCS, which is not achieved in the previous studies significantly.
computer science, artificial intelligence,engineering, electrical & electronic,operations research & management science