The success of a crowdsourcing application highly depends on whether a capable crowd can be recruited to undertake the corresponding tasks. For most applications, the cost is also a big concern. Recent research has been focused on incentive mechanisms for mobile crowdsourcing, which determine how to recruit a crowd mainly based on their prices/costs. However, scant attention has been paid to capability of the recruited crowd, i.e., the quality of services/data each individual mobile user and the whole crowd are potentially capable of providing. The project team aims to design a new lifestyle-aware approach for mobile crowdsourcing. The basic idea is to maintain a (relatively) stable set of mobile users and learn to gain a comprehensive view of their life patterns (i.e., their capabilities) such that once a mobile crowdsourcing task is received, a good crowd, whose capabilities well match the task, can be quickly found. The objective of this project to enable high-quality mobile crowdsourcing by designing a novel and holistic solution for crowd recruitment, which utilizes an energy-efficient framework for learning lifestyles of mobile users via smartphone sensing, and then employs lifestyle-aware algorithms and incentive mechanisms for crowd recruitment. To achieve the above objective, the proposed research is organized into three cohesive research thrusts:
Thrust 1 Energy-efficient Lifestyle Learning: We will develop a system for lifestyle learning, which uses a unified framework to learn lifestyle of mobile users by characterizing their behaviors and habits, and predicting their future activities. Moreover, we will improve its energy efficiency by developing lifestyle-aware algorithms to wisely control the sensor data collection procedure to reduce energy usages with guaranteed accuracy under the guidance of a lifestyle model.
Thrust 2 Lifestyle-aware Crowd Recruitment and Incentive Mechanisms: First, we will address the case where costs of mobile users are fixed and known beforehand by conducting a comprehensive study of Quality of Crowd (QoC) modeling for various applications and developing efficient and effective lifestyle-aware algorithms for crowd recruitment with the objective of maximizing QoC. Moreover, we plan to develop novel auction-based and fine-grained incentive mechanisms with provably-good QoC and nice auction-related properties (such as truthfulness and individual rationality).
Thrust 3 Implementation and Performance Evaluation: This task will be carried out in two steps. First, we will implement the proposed lifestyle learning system and conduct real experiments to validate and evaluate the proposed learning framework. Second, we will perform extensive simulation to evaluate the proposed algorithms and incentive mechanisms for crowd recruitment based on real data and lifestyle learning results in the first step.
Keywords: mobile crowdsourcing; smartphone sensing; energy-efficient control; incentive mechanisms.
This project is funded by NSF under grant #1525920.
Key Personnel:
- Jian Tang (PI)
- Jing Wang (Research Assistant)
- Lin Wu (Research Assistant)
- Kun Wu (Research Assistant)
- Zhiyuan Xu (Research Assistant)
- Chengxiang Yin (Research Assistant)