Networked Combinatorial Auction for Crowdsourcing and Crowdsensing

Yuhang Guo, Dong Hao, Mingyu Xiao, Bin Li

Abstract: We propose a novel protocol for crowdsourcing and crowdsensing by integrating combinatorial auctions and networks. With this protocol, agents who have already participated in crowdsourcing or crowdsensing tasks are incentivized to invite other agents to join the task. This new protocol aims to attract a growing number of skilled agents, thereby significantly improving the performance of crowdsourcing and crowdsensing. The challenge to this objective is how to incentivize each participant not only to contribute her full ability but also to try her best to spread the task information to her neighbors in the network. This problem is called networked auction or diffusion auction, which is a very new topic in algorithmic game theory and AI and has attracted considerable attention in recent years. Notably, current diffusion auctions lack customization for crowdsourcing or crowdsensing scenarios. Furthermore, no established diffusion auction adequately addresses the intricacies of handling multiple or even combinatorial tasks. This work is the first to design networked auction protocols for combinatorial tasks. From both theory and experiments, we have shown that the new protocols are proven to be incentive compatible, and that both the system’s cost and the requester’s costs could be decreased.