Abstract: Understanding and imitating human driver behavior has benefited for autonomous driving in terms of perception, control, and decision-making. However, the complexity of multi-vehicle interaction behavior is far messier than human beings can cope with because of the limited prior knowledge and capability of dealing with high-dimensional and large-scale sequential data. In this talk, I will present a novel concept with a hierarchical framework, called traffic primitives – the fundamental building blocks of driving behavior scenarios in a temporal space, which would be one of the ways to address tasks such as automatically understanding and modeling complex driving scenarios in a human-level semantic way without any prior knowledge required. In the second part of the talk, I will also give some applications on how Bayesian nonparametric statistics can be used to do driving behavior analysis and modeling, such as car-following driving style analysis and multi-vehicle interaction pattern learning at challenging scenarios.

Bio: Dr. Wenshuo Wang received the Ph.D. in Mechanical Engineering from Beijing Institute of Technology in June 2018. Before he joined the Safe AI Lab at CMU on July 2018, he was a Research Scholar in the Vehicle Dynamic & Control Lab (VDL) at University of California at Berkeley (UCB) from Sept. 2015 to Sept. 2017; and he was also Research Assistant in the Department of Mechanical Engineering at University of Michigan (UM), Ann Arbor from Sept. 2017 to July 2018. His research interests are human driver behavior modeling and prediction, human-vehicle shared control, multi-vehicle interaction decision-making, and Bayesian nonparametric learning and its applications. He received the Best Ph.D. Dissertation Award in China Society of Automotive Engineers (SAE-China) in 2018 and the Excellent Ph.D. Programs Foundation Award of Beijing Institute of Technology in 2017.

Homepage: https://sites.google.com/site/wenshuowangbit