1558-1748 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information. This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JSEN.2018.2873050, IEEE Sensors Journal 1 Real-time Vehicles Tracking Based on Mobile Multi-sensor Fusion Siim Plangi, Amnir Hadachi, Member, IEEE, Artjom Lind and Abdelaziz Bensrhair, Abstract—Tracking and positioning vehicles for navigation purposes are very important in many applications, ranging from simple routing problem to emergency planning and interventions. To this end, numerous approaches and industrial applications have been proposed to serve this purpose; however, there is a lot of work to be done to refine real-time systems for tracking in order to increase their accuracy and performance. Hence, in this paper we are presenting a real-time system that takes advantage of all the embedded sensors in the smartphone, using multi- sensor fusion to provide a real-time system capable of tracking, positioning, and even anticipating the maneuvers of vehicles in real-time. Index Terms—Tracking, Mobile sensors, Multi-sensor fusion, Kalman filter. I. I NTRODUCTION In recent years, we have witnessed an exponential increase in the use of smartphones as sensors [1] due to all the embedded sensors in it, such as GPS, gyroscope, etc. In addition, the marriage that occurred between the Internet and mobile technologies has significantly influenced the emergence of many applications and real-time systems for trackings, such as navigation systems, autonomous vehicles systems, location- based services for the smart city, urban planning, mobility analysis, and targeted marketing. Real-time tracking of vehicles has been of interest in many industrial products and research applications related to parking systems [2], racing and rally [3], and fleet management [4]. Therefore, the expectation behind this type of real-time sys- tems is the quick response and the quality of the positioning estimation and its parameters. However, this process can be interrupted due to any malfunction of the traditional GPS receiver or the embedded ones. Therefore, in this paper, we propose a method that takes advantage of the available sensors within the smartphone in order to refine the tracking using the multi-sensor fusion technique based on Kalman filter and map matching. The rest of the paper is organized as follows. The second section discusses the work related to the development of real- time systems for localization, tracking, and navigation. Then, the third and fourth sections present our proposed real-time system with all its algorithms and modules. Finally, the last two sections are about the results obtained and the future perspective that can be applied to make our system better. S. Plangi, A. Hadachi and A. Lind are with ITS Lab, Institute of Computer Science, University of Tartu, Tartu, 51014 Estonia e-mail: hadachi@ut.ee A. Bensrhair is with LITIS Lab, INSA de Rouen, Avenue de lUniversit, 76800 Saint-tienne-du-Rouvray, France II. RELATED WORK Being able to localize and track vehicles in real-time is a key builder in many applications, especially the ones related to smart transportation [5]. Therefore, providing precision for real-time systems in positioning and tracking vehicles is fundamental in intelligent transportation systems’ applications [6], [7]. Furthermore, due to the canyon effect and reduction of the satellite visibility in dense urban environments, it is challenging to perform a good positioning estimation and in tracking vehicles [8]. From this perspective, many researchers investigated possi- bilities to increase the accuracy in tracking and positioning vehicles in urban areas and it can be summarized into two main directions: the most dominant one focuses on building sophisticated algorithms to fix GPS errors and the second one relies on multisensor fusion. In [9], the authors proposed a robust algorithm for tracking using reckoning and strong tracking filter (STF). STF has the ability to provide a good estimation even when the precision of the system models is not obtainable. The filter itself uses an adaptive fading factor to regulate the gain matrix in real time in order to contain the orthogonal principle. The new approach found to outperform the Kalman filter in the following scenarios: an inaccurate system model, erroneous initial value, and abrupt change of states. Another proof of concept was demonstrated in [10], where the authors implemented a real-time system for tracking and positioning vehicles. Their system was composed of a GPS transmitting embedded module to interface within the vehicle, and a GSM (Global System for Mobile communications) for data transfer. Furthermore, for increasing the accuracy of the location data a Kalman filter was used for filtering and preprocessing. The outcome of the system was a reduction in the error from 43 meters to 15 meters. The usage of this type of systems is not only applied to navigation or fleet management purposes but also to other domains, for example in the security of vehicles; especially in developing anti-theft systems. One of the examples was presented in [11], where the authors designed and developed a real-time vehicles’ locking and tracking system based on GSM and GPS technology. Their method for localization and tracking was mainly focusing on using GPS positioning and the GSM triangulation methods. The contribution in this system can be resumed with the introduction of mobile communication as an embedded system and the whole system was designed in a single ship. Concerning techniques based on fusion, they are relying on optimizing the infer state of information from multiple sensors