The localization of autonomous vehicles requires, accurate tracking of its position and orientation in all conditions. As modern cities evolve localization would require a more precise accuracy that up to the level of centimetre and decimetre. One of the most crucial struggles in global positioning system and inertial navigation fusion is that the accuracy of the algorithm is reduced during GPS interruptions. In recent days bigdata, machine and deep learning offer great opportunities, especially for future smart and industrial 4.0 autonomous applications. This research programme is aiming to investigate and deploy machine and deep learning approach to improve and reach the level of reliability, accuracy and robustness required at low-cost GPS/IMU unit. The programme will also present a tracking platform solution that would compensates the issues of lack of accuracy in existing localization methods. The initial result of this ongoing programme is presented and reported in this paper. The paper also covers the research programme future development plans and milestones.