发展论坛 (Development Forums)

Development Forum 3: 

 7月27日13:30-15:30

 July 27, 13:30-15:30

Control Beyond AI

AI时代的控制

        Chair: Jifeng Zhang (Academy of Mathematics and Systems Sciences AMSS, CAS, China)

                                  Shaoyuan Li (Shanghai Jiao Tong University, China)

Panelists: Guangren Duan (Southern University of Science and Technology, China)

                 Changjun Jiang (Tongji University, China)

                 Xiaomin Hu (KTH Royal Institute of Technology, Sweden)

                 Haiping Che (Huawei Technologies, China)

                 Xiang Yin (Shanghai Jiao Tong University, China)

Title: AI时代的控制理论待何去何从？

Abstract: 上世纪60年代至本世纪初，现代控制理论得到了飞速发展，并同时在许多工程中得到了成功的应用。随着计算机的计算能力的大幅提升，AI技术得到了前所未有的广泛应用。今天的许多问题都可以依靠计算机的强大计算能力和神经网络逼近来完成，人们对于控制系统分析和设计的理论方法的需求在大幅减弱。在这种背景下，控制理论研究该何去何从？本报告对此汇报一下笔者的一些简单思考，抛砖引玉，与广大控制界学者商榷和共勉。

段广仁 中国科学院院士，CAA Fellow, IEEE Fellow, IET Fellow；南方科技大学自动化学科创始人，哈尔滨工业大学控制理论与制导技术研究中心名誉主任；国家自然科学基金委的创新群体、重大项目和基础科学中心项目负责人。现（曾）任中央军委科技委国防科技专家、国务院学位委员会第八届控制科学与工程学科评议组召集人、国家863计划专家组成员、教育部科技委信息学部委员、中国自动化学会常务理事等职。作为第一完成人获得国家自然科学二等奖2项，另获第四届中国青年科技奖、中国自动化学会控制理论专业委员会杰出贡献奖和全国优秀科技工作者称号；发表SCI论文360余篇，出版英文著作3部；培养全国优秀博士学位论文获得者2人，培养的博士生中已有学生成长为国家优青、IEEE Fellow、国家杰青和中国工程院院士。

Title: Some Thoughts on Computation, Intelligence and Control

Abstract: 1. Understanding of science and technology development process: Observe and understand the characteristics of the Material World, Information World and Intelligent World from the research object, method and form, especially the development process and thinking of Artificial Intelligence. 2. Network Transaction Risk Prevention and Control: (1) Establish the Behavior Theory of Concurrent Systems, which reveals the behavior correlation mechanism of system decomposition, and provides a theoretical basis for subsequent business crosscutting and behavior authentication. Propose the Moment Generating Function method of quantitative analysis, which provides optimization guidance for concurrent scheduling. (2) Invent the Concurrent Scheduling Technology of Business Process Network Crosscutting, propose the Flow Decomposition Dynamic Adaptation Method, solve the uncertainty and efficiency problems of concurrent computing resource allocation, reduce the transaction identification time from 200 milliseconds to 44.14 milliseconds, and form an independent innovation technologies in the field of Stream Computing in China. (3) Propose the Behavior Authentication Methods for Network Transaction Risk Prevention and Control, establish a layered risk control mechanism, subverted the traditional mode of transaction risk control, develop an network transaction risk control system, and increase the transaction recognition rate from 44% to 96.91%, leading the vigorous development of risk control in the network transaction industry.

Changjun Jiang Expert in computer field, academician of Chinese Academy of Engineering, Vice Chairman of Shanghai Association for Science and Technology. Professor of Tongji University, Honorary Professor of Brunel University London, Fellow of British Society of Engineering and Technology, National Outstanding Youth, Chief Scientist for National 973 Planning Program, National Outstanding Scientific and Technological Worker, winner of National Innovation Prize, etc. The achievements have won 106 invention patents from China, the United States, Germany and other countries, and 22 international PCTs; published more than 300 papers (including 82 ACM/IEEE Transactions) and 5 Chinese and English Monographs; won a second prize of National Technological Invention (Ranking 1), two second prizes of National Scientific and Technological Progress (Ranking 1), Ho pan Ching Yi Award (independent), etc.

He has been committed to the research of Network Computing for a long time. His relevant achievements have been positively evaluated and cited by academicians and other well-known experts from the United States, Britain, Germany, Sweden, India and other countries. He has presided over the establishment of China's first Network Transaction Risk Prevention and Control Architecture, System and Standard, and made pioneering contributions to China's becoming an International leader in this field.

Title: From Data-Driven Control to Emergence

Abstract: As machine learning is increasingly used in control systems, identifying system models and/or control laws from observed data, including data obtained from nature and human beings, becomes important. This is in particular critical for achieving emergences for multi-agent systems, especially non-consensus emergences, which is also connected to Distributed Artificial Intelligence (DAI). In this presentation we will discuss how tools such as non-cooperative differential games, (inverse) optimal control and manifold theory can be used in this context. The Swedish WASP program that focuses on AI and autonomous systems will also be briefly explained.

胡晓明亚历桑那州立大学获哲学博士学位,导师为著名非线性系统专家Christopher Byrnes。1989年获瑞典皇家理工学院（KTH）的Göran Gustafsson博士后奖金，1991年后在KTH数学系任职。2003 年晋升为优化与系统理论方向正教授。现任优化与系统理论实验室任，曾兼任KTH机器人中心副主任，及KTH网络系统国家重点中心（ACCESS Linaeus Center）的执委会委员，及该两中心董事会董事。胡晓明教授领导或参与了大量来自欧盟，瑞典研究基金会，瑞典战略研究基金会，后勤装备部及瑞典空间中心的科研项目，以及大量国际会议的组织工作，国际期刊的编委工作。研究方向主要为非线性反馈控制、非线性观测器设计、传感与主动感知、多自体系统的建模，分析及控制等，是目前活跃在国际控制理论领域的著名学者。他迄今发表200多篇论文及两本专著。

Title: 联合产业力量，共创AI时代新型工控

Abstract: 传统工控源自于西方的产业革命，将自动化技术运用在机械工业制造环节，实现自动加工和连续生产，提高机械生产效率和质量，在规模工业经济时代是释放生产力的主要作业手段。其典型特征是软硬件一体，生态垂直封闭，各自工控产品自成体系。传统工控经历了四个发展阶段，在工业4.0时代，随着人工智能、大数据、5G等新一代信息技术与制造业深度融合，原有的工控系统不能满足制造企业柔性生产需求，针对工业企业越来越复杂的远控、集控、现场控制等多场景控制需求，构建开放的广域协同大规模新型工控是行业趋势，以实现制造业行业知识的柔性生产，促进我国工业走向高端化、精细化、柔性化。

车海平  华为高级副总裁，数字转型首席战略官 1997年毕业于南开大学自动控制理论与应用专业，获博士学位，并加入华为。现任华为高级副总裁，数字转型首席战略官，负责华为在数字转型产业发展方面工作。与业界伙伴共同探索行业数字化共性规律和各自路径，牵引ICT产业转型，助力行业生产力提升。在华为，历任GSM核心网软件工程师、短消息中心开发经理、移动智能网PDT经理，WCDMA核心网PDT经理，移动核心网产品部部长，无线产品线副总裁，核心网业务部常务副部长，业务与软件产品线总裁，华为软件公司副总裁（技术，CTO），电信软件与核心网业务部副总裁&CTO等管理职位。因移动智能网技术创新和预付费业务的社会经济效益，于2001年荣获中国国家科技进步一等奖。积极参与数字转型相关的全球产业组织工作，目前担任TM Forum（电信管理论坛）董事、The Open Group董事、工业互联网产业联盟AII副秘书长、国家智能制造专家委员会委员、SAC/TC592主任委员等职务。

Title: Towards AI-Enabled Formal Control Theory in Cyber-Physical Systems

Abstract: Cyber-physical control systems are at the core of key infrastructure in our society. Ever-increasing demands for safety, security, and certification of these critical systems put stringent constraints on their design and necessitate the use of formal model-based approaches. This talk presents recently developed formal methodologies for analysis and control of cyber-physical systems. First, we focus on the design of safety-critical cyber-physical systems and we will present correct-by-construction design approach for controller synthesis. Then we will also discuss recent attempts towards secure-by-construction in cyber-physical systems. Finally, we will discuss recent developments in AI-enabled techniques for the formal design of cyber-physical systems.

Xiang Yin received the B.Eng degree from Zhejiang University in 2012, the M.S. degree from the University of Michigan, Ann Arbor, in 2013, and the Ph.D degree from the University of Michigan, Ann Arbor, in 2017, all in electrical engineering. Since 2017, he has been with the Department of Automation, Shanghai Jiao Tong University, where he is an Associate Professor. He received the 1000 young talents program in 2018. His research interests include formal methods, discrete-event systems and cyber-physical systems.

Dr. Yin is serving as the co-chair of the IEEE CSS Technical Committee on Discrete Event Systems, an Associate Editor for the Journal of Discrete Event Dynamic Systems: Theory & Applications, and a member of the IEEE CSS Conference Editorial Board. Dr. Yin received the IEEE Conference on Decision and Control (CDC) Best Student Paper Award Finalist in 2016.