Online auction for scheduling concurrent delay tolerant tasks in crowdsourcing systems

Abstract

Crowdsourcing is an emerging paradigm for human-powered problem solving in the Internet of Things (IoT) era. In this paper, we consider the online task allocation problem in crowdsourcing systems. Instead of considering real-time tasks that need to be completed immediately, we allow delay tolerant tasks which workers can complete at their own pace but within given deadlines. An online auction scheme (SOIL) is proposed for parallel allocation of delay tolerant tasks in practical crowdsourcing systems with stochastic requests from task owners and dynamic availabilities of workers. On one hand, SOIL provides a way for workers to evaluate their true cost for accepting new tasks with the given deadlines, and come up with optimal bidding strategies to maximize their utility. On the other hand, SOIL offers strategies for the system manager to design an online auction mechanism to maximize the system-wide utility and maintain a regulated crowdsourcing market. We prove that SOIL achieves a near-optimal system-wide utility with task deadline constraints. Moreover, SOIL is truthful, individually rational and budget balance. Through comprehensive simulations and rigorous theoretical analysis, we demonstrate the effectiveness of SOIL.