CO-ADAPT - Reliable Co-Adaptation for Connected Automated Vehicles

Despite many promises that connected automated vehicles (CAVs) would be imminent on our roads, it is widely recognized that the control algorithms supporting such technology are not mature. Indeed, control algorithms in CAVs only operate in linear regimes, being fragile to nonlinear operation like cooperative steering; tests show that CAVs are unable to interact safely on the road with human-driven vehicles, e.g., to form mixed platoons.

The CO-ADAPT project aims to develop a comprehensive theory of co-adaptation for CAVs, by leveraging control concepts of "nonlinear, robust, non-stationary adaptive control". CO-ADAPT will design nonlinear functionalities beyond mere linear regime; to achieve robust operation, control algorithms will be embedded with the capability to adapt inside non-nominal boundaries.

To promote co-adaptation on the road with human-driven vehicle, CO-ADAPT will design provably stable control algorithms that continuously adapt to the non-stationary behaviour of human drivers.

The adaptive control tools developed within CO-ADAPT will prove technical robustness and enhanced autonomy of CAVs in nonnominal traffic with nonstationary human-driving behavior. To target prompt technology transfer, CO-ADAPT will create an open source platform where the theory and adaptive control algorithms are verified through open-source autopilots and traffic simulators, via use cases like platoon merging and splitting, cooperative avoidance/lane change, mixed traffic with human-driven and automated vehicles.

CO-ADAPT will be carried out at Imperial College London under the supervision of Prof. Astolfi, a world-leading expert in nonlinear and robust adaptive control, nonholonomic mechanics, adaptive time-varying control. The project will involve a two-way knowledge transfer between the expertise of Prof. Astolfi and the expertise of the applicant in CAVs, cooperative adaptive cruise control, mixed traffic and autopilots.