大发棋牌www官方网址入口-主页欢迎您

12月29日:Dr. Xin Chen
发布时间:2021-12-28

报告题目Reachability Analysis of Learning-Enabled Cyber-Physical Systems

报告时间:12月29日 09:00

线下报告地点:理科大楼B1002

腾讯会议号:763-614-103

主持人:张民教授


报告摘要:

Neural networks have been increasingly used as the central decision makers in a variety of tasks. However, the use of neural-network controllers also gives rise to new challenges on verifying the correctness of the resulting closed-loop control systems especially in safety-critical settings. This talk focuses on computing the reachable sets of Cyber-Physical Systems (CPS) which are defined by neural network-controlled nonlinear continuous dynamics. Reachability analysis is the core of most formal verification tasks, and is notoriously difficult even on linear dynamics with simple switching logic. In the talk,  we will present the main framework of using flowpipe construction for learning-enabled CPS. Three techniques will be introduced for computing higher-order overapproximations for reachable sets. We will show how the dependencies among the state variables during the system evolution can be accurately approximated by our methods. Experimental evaluations will also be given to demonstrate the effectiveness of the methods.



报告人简介:

Dr. Xin Chen is currently an assistant professor at the University of Dayton in Dayton, Ohio. He received his Ph.D. in Computer Science from RWTH Aachen University in 2015, and was a postdoctoral research associate at the University of Colorado, Boulder. Dr. Chen's main research interests lie in the area of solving verification and synthesis problems for autonomous systems using formal methods. He is also the sole-PI of an on-going research project funded by the U.S. Air Force Research Laboratory.



华东师范大学大发棋牌www官方网址入口

www.sei.ecnu.edu.cn Copyright Software Engineering Institute

办公邮箱:office@sei.ecnu.edu.cn | 院办电话:021-62232550 | 学院地址:上海中山北路3663号理科大楼