The automotive sector is one of the richest targets for emerging innovations in Cyber Physical Systems (CPS). Our Research Laboratory has focused for the past ten years on advanced automotive safety systems.
The current research agenda is on robust certification of advanced autonomous and semi-autonomy systems. Instead of totally disconnecting the driver from the vehicle, as in autonomous cars, we envision a vehicle where the degree of autonomy is continuously changed in real-time as a function of certified uncertainty ranges in driver behavior and environmental uncertainties. The degree of autonomy is chosen among a continuum of options between the driver in total control of the vehicle and autonomous driving, depending on the trust level in the CPS. This human-centric autonomy has a high probability of societal impact (insurance policy will be the same if not lower, system will cost less and the market penetration higher). It will also have a direct impact on a large number of CPS with humans in the loop, which do require robustness certificates.
With this in mind, we propose a paradigm shift which looks at the whole cyber physical vehicle/environment/driver and thus addresses all its three main critical components: (A) the vehicle/environment interaction, (B) the driver uncertainty and (C) the provably-safe intervention under the predicted uncertainty of A and B. At Berkeley we are developing a novel science for Cyber-Physical Systems with the goal of obtaining a provably safe human-centric autonomy where certification is evidence-base and evolves with the system (as new driver behaviors, scenes, slipping dynamics are updated in the CPS database). Robustness is measured against bounded state-dependent uncertainty quantified from developed driver/vehicle/environment interaction models, calibrated and validated on large data set and updated in real-time.