Research And Design Of Volunteer Sensing Based Real-time Traffic Information Collection System

As a vital building block in the smart transportation system, realtime transportation data acquisition has draw much attention for its importance and value in applications and great potential in the coming ’big data era’. Mobile sensing based data acquisition became a hot topic for its low-cost, wide coverage and the rising population of smartphones and mobile internet devices, resulting in many research findings.In this paper, we studied the realtime transportation data acquisition based on mobile sensing. Combining traditional mobile sensing framework and volunteering computing, we propose the concept of volunteering sensing. The framework and working procedure of this volunteering sensing based realtime transportation data acquisition are designed via deep researching and mining of system and volunteer requirements. The findings are successfully fed back in real systems. The main works and contributions are as follows:Firstly, we briefly introduce the concept and function of the realtime transportation data acquisition sub-system. By comparing existing solutions in this field, we show the advantages and significance of using mobile sensing, and summarize its current inadequacies; Secondly, we introduce the main procedures of the system. Some important techniques, including localization, map matching and transpotation parameter estimation, are introduced and compared in order to justify our technique, algorithm and model selection in this paper. Considering the current inadequacies of mobile sensing, we introduce the essence of volunteering computing. The design and implementation of this system is given after in-depth system requirements analysis, along with two kinds of sensing strategies:accurate sensing and low-power sensing.Accurate sensing are designed for unconstrained energy supply conditions. By using geohash fast initialization algorithm, we minimized human intervention and supervision. Meanwhile, the high-complexity procedures, like map matching, are allocated to the smart devices to reduce computation loads in the cloud servers, improving its realtime performances. Low-power sensing are design for those energy-sensitive volunteers. Using machine learning technique, we propose to use the acceleration sensor for transportation state estimation, reducing the usage of GPS and other locating sensors, which are the main energy consumers in this system.A series of experiments have been performed to verify the performance of accurate sensing and low-power sensing. Result shows that accurate sensing strategy can provide the average speed and driving time of volunteers accurately while the low-power strategy can achieve90%accuracy.