Team Recruitment of Collaborative Crowdsensing: A Graph-Based Approach
Hui Liu, Xiaowan Li, Chuang Zhang, Xiaodong Chen, Weipeng Tai · IEEE Access · 2024
Collaborative crowdsensing is a teamwork approach that leverages the collective intelligence of participants. Unlike traditional crowdsensing user recruitment, collaborative crowdsensing not only considers users’ intrinsic abilities (sensing ability, reliability) but also places a greater emphasis on their extrinsic abilities(collaborative abilities). Initially, we calculated user’s collaborative ability based on an undirected graph, which is modeled by user’s interaction relationships and collaborative characteristics. Besides, a directed graph representing evaluation relationships is constructed through user ratings, and the PageRank algorithm is utilized to assess a user’s reliability. Taking into account users’ intrinsic abilities, extrinsic capabilities and cost quotations, we propose a team recruitment mechanism. which transforms the recruitment problem into a maximum spanning tree problem in graph theory. Furthermore, an enhanced Prim algorithm is utilized to select efficient users. To address the issue of recruiting new users lacking prior knowledge, an improved UCB algorithm is introduced on top of the proposed team recruitment mechanism. Additionally, to prevent unfair payments resulting from selfish users misreporting costs, the team recruitment process is modeled as a reverse auction, and a reward payment method based on Myerson theory is designed. This method ensures fair compensation based on a user’s utility contribution to the team. Finally, experimental comparisons on real datasets validate the effectiveness of the algorithms.