Evaluation of smart-phone performance for real-time traffic prediction

Smartphones efficiently collect traffic information, guide the drivers and inform the end-user about the current and future traffic conditions. Due to enhanced sensor technology, visualization capabilities, navigational performance and network connectivity, smartphones play an important role in advanced travel information systems (ATIS). Although they offer increasing computation power nowadays, this potential smartphone' resource has not been explicitly evaluated for neither ATIS nor Intelligent Transport System (ITS) applications. In this study, we actively involve the smartphone into real-time compressed prediction of large traffic networks. More precisely, we run prediction algorithms on the central server to obtain future state for the subset of the links in the network that we refer to as compressed network state. Then, we send the predicted values for compressed network state to smartphones where network extrapolation is performed. Network extrapolation involves vector-matrix multiplication where row vector represents the compressed network state while the matrix is stored on the mobile phone and contains the relationships function between the compressed state and entire network. Such decentralized infrastructure can significantly reduce the overhead of the communication network and enhance the development of cooperative, peer to peer networks for the NextGen Intelligent Transportation Systems applications.