Title: Towards Foundation Models for Intelligent Embedded Sensing

Abstract: As embedded computing increases its reliance on AI, a key challenge in building modern embedded applications becomes in the high cost of acquiring labeled training data. To reduce this cost, this talk discusses advances in exploiting unlabeled training data instead. The exploitation of unlabeled data at scale for AI training has been very successful in other contexts, such as pre-training of Large Language Models (e.g., ChatGPT) and Visual Language Models. However, the pretraining infrastructure for models in the natural language and vision domains is not necessarily best-suited for data modalities common to embedded Edge AI applications, such as numeric time-series data. Similarly, the tasks executed by embedded Edge AI applications often differ from those expected of Large Language Models. The talk discuses recent work on adapting self-supervised pre-training solutions originating from the language and vision domains to the needs of intelligent embedded applications, as well as initial evaluation of the efficacy and robustness of such adapted solutions at executing CPS/IoT tasks. We show how new foundation models can be developed for the CPS/IoT space that improve the efficacy and robustness of Edge AI, while minimizing the need for labeled data. Talk concludes with further opportunities and challenges in developing foundation models for intelligent embedded sensing.

Bio:Tarek Abdelzaher received his Ph.D. in Computer Science from the University of Michigan in 1999. He is currently a Professor and Willett Faculty Scholar at the Department of Computer Science, the University of Illinois at Urbana Champaign. He has authored/coauthored more than 400 refereed publications in real-time computing, CPS/IoT, distributed systems, intelligent networked sensing, machine learning, and control. He served as Editor-in-Chief of the Journal of Real-Time Systems for 15 years, and as Associate Editor of the IEEE Transactions on Mobile Computing, IEEE Transactions on Parallel and Distributed Systems, IEEE Embedded Systems Letters, the ACM Transaction on Sensor Networks, ACM Transactions on Internet Technology, ACM Transactions on Internet of Things, and the Ad Hoc Networks Journal. He chaired (as Program or General Chair) several conferences in his area including RTAS, RTSS, IPSN, Sensys, DCoSS, ICDCS, Infocom, and ICAC. Abdelzaher's research interests lie broadly in understanding and influencing performance and temporal properties of networked embedded, social, and software systems in the face of increasing complexity, distribution, and degree of interaction with an external physical environment. Tarek Abdelzaher is a recipient of the IEEE Outstanding Technical Achievement and Leadership Award in Real-time Systems (2012), the Xerox Award for Faculty Research (2011), as well as several best paper awards. He is a fellow of IEEE and ACM.

Keynote 2: 9:30-10:20 AM, Jiannong Cao (host: Guoliang Xing)

Title: Collaborative Edge AI Acceleration

Abstract: Edge AI is essential to support applications requiring low processing delays. However, existing edge AI approaches are inadequate to meet the demands of emerging advanced applications, such as autonomous vehicles, industrial IoT, digital twins, and VR/AR, which require ultra-low latency, large-scale deployment, and more powerful AI support. Most existing works focus on edge AI acceleration using vertical collaboration between edge and cloud. In this talk, I introduce a systematic approach to exploring heterogeneity and collaboration among edge devices for further acceleration of AI tasks. I will describe the challenging issues, including diversity of AI models, large scale and dynamicity of edge networks, and heterogeneity of edge devices, and propose cross-layer solution integrating hardware abstraction, resource management, task scheduling, and application execution throughout the hardware, software, and application layers.

Bio:Dr. Cao is the Otto Poon Charitable Foundation Professor in Data Science and the Chair Professor of Distributed and Mobile Computing in the Department of Computing at The Hong Kong Polytechnic University. He served the department head from 2011 to 2017. Dr. Cao is also the Dean of Graduate School, the director of Research Institute for AIoT and director of University’s Research Facility in Big Data Analytics. Dr. Cao’s research interests include edge computing and distributed systems, wireless sensing and networking, big data and AI. He published 6 co-authored and over 500 papers in major international journals and conference proceedings. He also obtained 13 patents. He received many awards for his outstanding research achievements. Dr. Cao served the Chair of the Technical Committee on Distributed Computing of IEEE Computer Society 2012-2014. He is a member of Academia Europaea, a fellow of HK Academy of Engineering Sciences, a fellow of IEEE, a fellow of CCF and a distinguished member of ACM.

Panel 1: 10:20-11:20 AM “Challenges and Opportunities for Next Generation AIoT”

Chair: Xiaofan (Fred) Jiang

Panelist: Rong Zheng, Wen Hu, Nan Guan, Jie Liu

Invited speech 1: 11:20-11:40 AM, Dongyao Chen

Title: Revealing Fine-grained Information with Magnetic Sensing

Invited speech 2: 11:40-12:00 AM, Xianjin Xia

Title: LoRa for Massive IoT Connections: Are We There Yet?

Lunch break: 12:00 AM 2:00 PM

Keynote 3: 2:00-2:50 PM, Naijun Zhan

Title: Reset Controller Synthesis

Abstract: Correct-by-construction controller synthesis provides a mechanism to guarantee the correctness and reliability of hybrid systems (HS) by design. Typical examples include reset controllers, feedback controllers, and switching logic controllers. Reset controllers steer the behavior of HS to attain system objective through restricting its initial set and redefining the reset map associated with discrete jumps, which is less explored in the literature but is of theoretical and practical significance. In this talk, I will summarize our recent work on the reset controller synthesis for HS that guarantee safety and liveness, in terms of transverse and generalized reach-avoid sets. The basic idea is to reduce the problem of guaranteeing safety and liveness properties to differential invariant generation and generalized reach-avoid problems. For polynomial hybrid systems, those problems can be solved by further reduced to convex optimizations. Moreover, changing a continuous evolution abruptly is counter-intuitive, even impossible, for example, it is impossible to change the velocity of a train from 0 km/h to 300 km /h instantaneously in reality, although it is mathematically simple. So, realistically, we should consider this issue in the context of time-delay Thus, we investigate the reset controller synthesis problem for delay hybrid systems (dHS), which contains delay in both continuous evolution and discrete transitions and propose a novel reach-avoid analysis based method.

Bio:Naijun Zhan is a Boya distinguished professor in the School of Computer Science of Peking University. He got his bachelor degree and master degree both from Nanjing University, and his PhD from ISCAS. Prior to join Peking University, he worked at the Faculty of Mathematics and Informatics, Mannheim University, Germany as a research fellow, and afterwards worked at Institute of Software Chinese Academy of Sciences (ISCAS) as an associate professor, a full professor, and a distinguished professor. His research interests cover formal design of real-time, embedded and hybrid systems, program verification, modal and temporal logics, and so on. He is in the editorial boards of Journal of Automated Reasoning, Formal Aspects of Computing, Journal of Logical and Algebraic Methods in Programming, Journal of Software, Journal of Electronics, and Journal of Computer Research and Development and so on, a member of the steering committees of SETTA and MEMOCODE, the pc co-chairs of FM 2021, SETTA 2016, the general co-chairs of MEMOCODE 2018, MEMOCODE2019 and ICESS 2019, and serves more than 100 international conferences program committees e.g., CAV, RTSS, HSCC, FM, TACAS, EMSOFT and so on. He published more than 140 papers in international leading journals and conferences, 2 books and 4 book chapters, and edited 4 conference proceedings and 7 journal special issues. See lcs.ios.ac.cn/~znj for more details.

Keynote 4: 2:50-3:40 PM, Lili Qiu

Title: The Future of Healthcare Powered by AI & Wireless Sensing

Abstract: The multifaceted nature of individual health can be captured using an array of physiological and behavioral indicators, including but not limited to respiratory patterns, cardiac rhythms, neural activity as evidenced by brain waves, articulation and linguistic nuances, kinesthetic dynamics, and the intricate details captured in medical imaging. This talk will present innovative wireless sensing and advanced machine learning technologies, which have the potential to transform daily health monitoring and revolutionize the accuracy and efficiency of disease diagnosis. Furthermore, the talk will share the valuable lessons we have learned through our collaboration with leading hospitals.

Bio:Dr. Lili Qiu is Assistant Managing Director of Microsoft Research Asia and is mainly responsible for overseeing the research, as well as the collaboration with industries, universities, and research institutes, at Microsoft Research Asia – Shanghai. She obtained her MS and PhD degrees in computer science from Cornell University. Dr. Qiu is an expert in Internet and wireless networking. In 2005, she joined the University of Texas at Austin as an assistant professor in the Department of Computer Science, and later, in view of her outstanding achievements in the internet and wireless networks fields, she was promoted to a tenured professor and doctoral advisor. Dr. Qiu is an IEEE Fellow, a NAI Fellow and an ACM Fellow and also serves as the ACM SIGMOBILE chair. She was named an ACM Distinguished Scientist and was a recipient of the NSF CAREER award, among many other honors.

Panel 2: 3:40-4:40 PM “大模型时代的嵌入式智能“

Chair: Bin Guo

Panelist: Mo Li ，Yanyong Zhang, Chenren Xu ，Qian Zhang , Wenbo He

Invited speech 3: 4:50-5:10 PM, Chenshu Wu

Title: In-Cabin Automotive AI via Statistical Acoustic Sensing

Invited speech 4: 5:10-5:30 PM, Sicong Liu

Title: Enabling Resource-efficient mobile and embedded System with cross-level optimization