Invited Speaker: Professor Huei Peng
             Department of Mechanical Engineering, University of Michigan

Abstract

Many driver assistance systems (DAS) were proposed over the last decade to enhance the comfort, safety and efficiency of ground vehicles. Examples include adaptive cruise control (ACC), vehicle stability control (VSC) systems, Collision Warning/Collision Avoidance, etc. As oppose to fully automated designs, for driver assistance systems, the human driver retains ultimate authority and thus the responsibility of safe driving.

Since these DAS always work with a human driver co-existing, the overall vehicle performance will depend on, not how well the DAS works by itself, but rather its interaction with the human driver. The evaluation of DAS, therefore, needs to be conducted with human (driver) in the loop. Up to recently, this mostly means testing on prototype or experimental vehicles, or high-fidelity driving simulators, which are all costly and time-consuming. The alternative, which relies on simulation based product development implies the need for an accurate human driver model. This requirement is non-trivial because of the time-varying and uncertain nature of human drivers. Furthermore, human decisions are complex and involve pre-cognitive, planning-strategic as well as servo-regulation actions, and the switching among different goals, as well as sensory cues. In addition, for certain driving tasks, human drivers employee preview-predictive strategies which are usually difficult to implement. Finally, human perception and decision process were sometimes based on fusion of complex visual and motion feedbacks that are difficult to understand and reproduce in mathematical models.

The focus of this talk is on the evaluation, and development of driver models using naturalistic driving data, with the ultimate goal of helping the design and evaluation of active safety systems. The driving data are based on several databases recently constructed at the University of Michigan Transportation Research Institute (UMTRI): The SAVME database, the ICC FOT database, and the ACAS database. Several examples on using these driving data for the design and evaluation of collision warning/avoidance systems, as well as the development of “errable” driver model are discussed.

About the Speaker:

Dr. Huei Peng is a Professor in the Department of Mechanical Engineering (2005) of the University of Michigan, where he has been a faculty since 1993. He has been the Director of College of Engineering Automotive Engineering Program (April 2002-2007), and currently also serves as the Executive Director of Interdisciplinary and Professional Engineering (2007-present). Earlier, during 1992-93 he was an assistant engineer at the University of California Berkeley, PATH Program. Prof. Peng’s areas of research are adaptive control and optimal control, with emphasis on their applications to vehicular and transportation systems. Current thrusts of his research include modeling and design of hybrid vehicles and fuel cell control systems, and active safety systems for ground vehicles. His work is well documented in 120+ technical publications, including 50 journal papers and two books. He has served as PI or co-PI for 28 funded research and education projects with a total budget of more than 3 million dollars. He has held leadership roles in ASME, and has provided significant services to NSF and various journals and conferences as panel reviewer, editor, organizer, and chair.

Location: CEE Conference Room, Phillips Hall 6^th Floor
          The George Washington University, 801 22nd St. NW, Wash. DC 20052
Time: Friday, April 27, 2007, 10:45 AM-12:00 PM