A cooperative sensing and mining system for transportation activity survey

This paper exploits smartphones to design a transportation activity survey system that investigates when, where and how people travel in an urban area. In such a system, the essential requirement is collecting and processing big data which will raise two critical issues, energy-conservation and scalability. To address the former issue, the GPS sleeping interval of a smart-phone is controlled by the back-end servers adaptively based on the real-time moving speed and transportation modes. To address the latter issue, we consider MapReduce to design the back-end Cloud, where intelligent learning and classification algorithms are implemented to detect the stops and transportation modes and provide smartphones with an appropriate GPS sleeping interval based on the GPS statistics on the back-end Cloud. The unique feature of our system is to integrate participatory sensing and Cloud-enabled processing system closely which incorporates knowledge extracted from the Cloud (i.e., transportation modes) into sensing control of smartphones. In this way, sensing control could be optimized through the knowledge behind crowdsourced data. Our system has been deployed in Singapore to support the Land Transport Authority´s transportation activity survey over 1 year. Extensive experimental results indicate that our system can reduce the energy consumption of smartphones efficiently and process concurrent data arrival from a huge number of users.