NSF-CPS: Adaptive Intelligence for Cyber-Physical Automotive Active Safety System Design and Evaluation

The main objective of this research is to use techniques and models from human factors, computational neuroscience, and adaptive and real-time optimal control theory in order to investigate the effects of the introduction of learning and adaptation to the next generation of ASCS. In particular, we will:
(a) Learn the driver’s habits, driving skills, patterns and weaknesses.
(b) Model his/her current cognitive state along multiple dimensions such as attentiveness, aggressiveness, etc.
(c) Use this information in order to adapt the ASCS to the particular situation at hand to achieve maximum
performance in terms of safety and comfort.
(d) Evaluate the theoretical developments via a comprehensive testing plan.
We will focus mainly on the driver’s state estimation problem, based on his/her attention from visual data, along with his/her steering, braking, and accelerating commands while driving. The outputs of the estimation layer (based on probabilistic graphical models) are parameters (or their suitable surrogates) of control-theoretic models, similarly to those commonly used to model driver-vehicle interaction. However, we will go one step further: we will use these parameters to design and adapt the ASCS in real-time. The novelty here is the on-the-fly, continuous adaptation of the ASCS parameters as they become available from the driver estimation models.

Map of Cognitive Engineering Center

Cognitive Engineering Center (CEC)
Georgia Institute of Technology
270 Ferst Drive
Atlanta GA 30332-0150