Smart, Connected and Autonomous Vehicle and Energy systems for efficient, safe, secure, and sustainable transportation in metropolitan cities

Connected and Autonomous Transportation (CAT) is viewed as the future of all transportation because of its promise to ensure increased efficiency and safety in city traffic as well as on highways. On the other hand, for CAT to be acceptable by the society as a new transportation model, all of its challenges and needs should be adequately met; i.e. safe and efficient performance under dynamical/environmental uncertainties, mitigation and resilience against security threats resulting from connectivity, as well as very fast DC charging station (FDCS) for increased CAT efficiency, which calls for high energy levels and supporting microgrids. The connectivity inherent for CAVs result in an interaction between all subsystems; i.e. between the vehicles while in action, and between vehicle, FDCS, and grid during charging, thus increasing the attack surfaces and dependence of the overall CAT on the performance of its subsystems and their interactions. While this interdependence is a well-recognized fact, the existing literature mostly reports studies limited to certain functions of CAT, often evaluated with lab experimentation only and by ignoring the interaction of subsystems/functions. This proposal considers transportation and energy systems to be integral and critical components for CAT, takes a learning-based approach to build reliable autonomy, and recognizes real-world experimentation to be essential for the adequate evaluation of the schemes and their interaction. This is one of the important novelties of this project, with strong potential to address a major challenge that the AI industry faces today, namely, dealing with the integration and adaptation of the ‘closed-world’ laboratory solutions into a more volatile and unpredictable ‘open-world’. To this aim, the proposed CAT will not only take into consideration the development of a CAV system with Level-4 autonomy, but also a microgrid of mostly renewable energy sources supporting a FDCS, and smart energy management systems (EMS) that together ensure the sustainability of CAT within a real-world scenario. To address the many uncertainties involved with the real-world CAT implementation, in this project, we will focus on developing novel Deep/Reinforcement Learning (D/RL) based modular schemes for each autonomy function of the vehicle; i.e. perception, localization, motion planning, and tracking control, as well as for the autonomous Energy Management Systems (EMS) also taking security measures by integrating ML based anomaly detection and Moving Target Defense strategies into the developed schemes. All autonomy functions will be tested and evaluated in integration at the ITU Living Lab, with two Level-4 autonomous Smart Cars and one autonomous KARSAN bus to perform an autonomous ring service around campus, supported by an FDCS, which, in turn, will be supplied by a microgrid consisting of renewable energy sources, namely, solar panels and a recuperation system to capture the regenerative braking energy released from the metro cars arriving at the ITU Maslak Metro Station. The project will specifically contribute to the design of novel RL schemes for real world challenges, which will be evaluated for an integrated, fully autonomous transportation and energy system in a real-world setting. This is novel not only for Turkey, but for the world, will fill an important research gap for RL based autonomy. We also expect the proposed first-and last-mile pilot implementation at the ITU campus in collaboration with our industrial partners, Metro Istanbul, ADASTEC, and ENTES to contribute to the acceptability of the CAT technologies by society, further motivating the use of public transportation in metropolitan cities. Within the planned bilateral collaboration, HBKU will provide expertise in all ML aspects of the project, particularly in the development and integration of anomaly detection schemes into the CAT system to ensure its safety and security, and development of deep NNs for all the schemes.

Hamad Bin Khalifa University (HBKU)