Tuesday, July 18, 2017 - 2:15pm
3110 Etcheverry Hall
Dr. Kang Li & Dr. Chi-Sheng Shih
National Taiwan University
TALK #1: 2:15 - 3:00 PM
"Model-based Control System Design and XiLS Validation Techniques for Innovative Electric Propulsion and Automated Driving System Development"
Dr. Kang Li
Abstract: Electrification, automation and connectivity are the trends of future automotive technology development. Cross-disciplinary technologies including vehicular/power electronics, wireless communication, sensors, advanced control techniques and artificial intelligence (A.I.), etc., are applied and integrated in developing modern automotive control systems. To ensure the success of development and implementation of novel EV/AV control systems, a systematic approach for automotive control system fast prototyping, testing, and validation is needed. In this talk, a unified automotive control system development methodology, which combines the model-based control design and X-in-the-loop simulation (XiLS) techniques, will first be presented. Then, examples of automated and electric vehicle control system development using this methodology will be given. Finally, the on-going research on further incorporating A.I. technology into the proposed XiLS framework will be described.
Biography: Dr. Kang Li is Associate Professor of the Department of Mechanical Engineering, National Taiwan University (NTU), the Director of Intelligent Vehicles & Mechatronics Laboratory, NTU and the Deputy Director of New Energy Center, NTU. He has a wide spectrum of research interests including vehicle electrification and automation, mechatronics, telematics and solar energy systems. He earned his B.S. degree in Mechanical Engineering from NTU in 2000, and M.S. and Ph.D. degrees in Mechanical Engineering from the University of California-Berkeley in 2016 and 2009, respectively.
TALK #2: 3:00 - 3:45 PM
"Federating Edge and Cloud Devices for Deep Learning on Automotive Applications"
Dr. Chi-Sheng Shih
Abstract: Object detection is one of the most fundamental functions for many automotive applications. Given the rigorous safety regulations, it is required to provide accurate detection results in real-time using limited computation resources. Deep learning approach has shown its promising capability on detecting objects in images and videos under highly dynamic environments. However, it requires great amount of computation resources and has limited use in automotive applications. To overcome the needs on computation resources for deep learning and tackle the accuracy requirements for automotive applications, the talk will discuss the study on federating edge and cloud devices to enable real-time object detection at high accuracy. Compared to isolated embedded platforms, the proposed architecture can increase the detection accuracy up to 80x and increase the detection frame rate up to 10x at the same time. This approach is not limited to any specific deep learning engine.
Biography: Dr. Chi-Sheng Shih has joined the Department of Computer Science and Information Engineering at National Taiwan University on Feb. 1, 2004. He currently serves as an Associate Professor. He received the B.S. in Engineering Science and M.S. in Computer Science from National Cheng Kung University in 1993 and 1995, respectively. In 2003, he received his Ph.D. in Computer Science from the University of Illinois at Urbana-Champaign. His main research interests include embedded systems, hardware/software codesign, real-time systems, and database systems.
Hosted by: Professor Masayoshi Tomizuka, 5100B Etcheverry Hall, 510- 642-0870, email@example.com &
Dr. Ching-Yao Chan (California PATH)