Finish Them on the Fly: An Incentive Mechanism for Real-Time Spatial Crowdsourcing

Abstract

Proper incentive mechanism design for stimulating workers is a fundamental challenge in nowadays spatial crowdsourcing (SC) powered applications like Didi and Uber. Usually, extra monetary rewards are paid to workers as incentive to enhance their participation in the SC platform. However, deciding incentives in real-time is non-trivial as the spatial crowdsourcing market changes fast over time. Existing studies mostly assume an offline scenario where the incentives are computed considering a static market condition with the global knowledge of tasks and workers. Unfortunately, this setting does not fit the reality where the market itself would evolve gradually. In this paper, to enable online incentive determination, we formulate the problem of Real-time Monetary Incentive for Tasks in Spatial Crowdsourcing (MIT), which computes proper reward for each task to maximize the task completion rate at real time. We propose a unified and efficient approach to the MIT problem with a theoretical effectiveness guarantee. The experimental results on real ride-sharing data show that, compared with the state-of-the-art offline algorithms, our approach decreases the total worker response time by two orders of magnitude with insignificant utility loss.