Federal University of Rio Grande do Sul, Brazil
In the past years, several works have proposed custom hardware and software-based techniques for the acceleration of Convolutional Neural Networks (CNNs). Most of these works focus on saving computations by changing the used precision or modifying frame processing. To reach a more aggressive energy reduction, in this talk we discuss how the underlying hardware dissipates energy, and show how software-only modifications to the CNNs inference can improve the process. Our approach exploits the inherent locality in videos by replacing entire frame computations with a movement prediction algorithm. Furthermore, when a frame must be processed, we avoid energy-demanding floating-point operations, and at the same time reduce memory accesses by employing look-up tables in place of the original convolutions. Using the proposed approach, one can reach significant energy gains of more than 25× for security cameras, 12× for moving vehicles applications, and up to 96x for complex and deep architectures.
''Dr. Luigi Carro received the Electrical Engineering and the MSc degrees from Universidade Federal do Rio Grande do Sul (UFRGS), Brazil, in 1985 and 1989, respectively. From 1989 to 1991 he worked at ST-Microelectronics, Agrate, Italy, in the R&D group. In 1996 he received the Dr. degree in the area of
Computer Science from Universidade Federal do Rio Grande do Sul (UFRGS), Brazil. He is presently a full professor at the Applied Informatics Department at the Informatics Institute of UFRGS, in charge of Computer Architecture and Organization. He has advised more than 20 graduate students, and has published more than 150 technical papers on those topics. He has authored the book Digital systems Design and Prototyping (2001-in Portuguese) and is the co-author of Fault-Tolerance Techniques for SRAM-based FPGAs (2006-Springer), Dynamic Reconfigurable Architectures and Transparent optimization
Techniques (2010-Springer) and Adaptive Systems (Springer 2012). In 2007 he received the prize FAPERGS - Researcher of the year in Computer Science. His most updated resume is located in Lattes. For the latest news, please check webpage.''
University of Münster, Germany
Embedded systems have become ubiquitous in our daily lives, and their complexity continually evolves to unprecedented levels. In addition to their heterogeneity and interaction with a physical environment, we see a tremendous increase in the use of learning to make autonomous decisions in dynamic environments. These developments pose significant challenges for ensuring the safety and reliability of embedded systems. Formal methods have the potential to guarantee crucial safety properties under all circumstances, but are incredibly expensive and severely suffer from scalability issues. In this talk, I will summarize some of our recent efforts towards more scalable verification of embedded control systems. Our main contributions are novel encodings, contracts for learning, and reusable formalizations, which enable us to leverage deductive verification techniques to heterogeneous hardware/software systems and intelligent hybrid control systems.
Dr. Paula Herber is a full professor and head of the Embedded Systems Group at the Computer Science Department at the University of Münster, Germany. Since May 2021, she is also a part-time full professor at the University of Twente in the Formal Methods and Tools (FMT) group in the Netherlands. She has received her Ph.D. from TU Berlin in 2010, and worked as a postdoc at the International Computer Science Institute (ICSI) in Berkeley, California, as a substitute professor at the University of Potsdam, and as a postdoc and independent research group leader in the Software Engineering and Embedded Systems group at TU Berlin. Her main research interests are quality assurance for embedded systems, test automation, and formal methods. She is best known for her contributions to the formalization of industrially used system design languages such as SystemC and Simulink, and highly interested in new techniques to increase the applicability of formal methods for embedded and cyber-physical systems.
Dresden University of Technology (TU Dresden), Germany
The complexity and high demand for real-time and energy-efficient computing for autonomous robots and drones, requires novel runtime adaptive computing systems. Data from multiple sensors must be processed in parallel with a variety of signal/image processing and machine learning algorithms. In addition, the autonomous robot must quickly adapt to changing situations at runtime, e.g., by switching the executed algorithm from navigation to object detection. To achieve high energy efficiency, this adaptation requires a change in both the signal/image processing algorithms and the underlying computing architecture for these specific algorithms.
This talk will present concepts and realizations for such an approach, consisting of an adaptive domain-specific computing architecture and its design and programming methodology. The importance of such an approach will be demonstrated using several research projects with robotics and drone applications.
Dr.-Ing. Diana Göhringer is professor for adaptive dynamic systems at TU Dresden, Germany. From 2013 to 2017 she was an assistant professor and head of the MCA (application-specific Multi-Core Architectures) research group at the Ruhr-University Bochum (RUB), Germany. Before that she was working as the head of the Young Investigator Group CADEMA (Computer Aided Design and Exploration of Multi-Core Architectures) at the Institute for Data Processing and Electronics (IPE) at the Karlsruhe Institute of Technology (KIT). From 2007 to 2012, she was a senior scientist at the Fraunhofer Institute of Optronics, System Technologies and Image Exploitation IOSB in Ettlingen, Germany (formerly called FGAN-FOM). In 2011, she received her PhD (summa cum laude) in Electrical Engineering and Information Technology from the Karlsruhe Institute of Technology (KIT), Germany. She is author and co-author of over 150 publications in international journals, conferences and workshops. Additionally, she serves as technical program committee member in several international conferences and workshops (e.g. DATE, ICCAD, FPL). She is reviewer and guest editor of several international journals. Her research interests include reconfigurable computing, Multiprocessor Systems-on-Chip (MPSoCs), Networks-on-Chip, Hardware-Software-Co-design, simulators/virtual platforms and runtime systems.
Communication-Computing-Control Co-Design for the Cloud Fog Automation: a New Paradigm of Realizing Industrial Automation Systems
Senior Principal Scientist, ABB Corporate Research Sweden
Adjunct Professor, KTH Royal Institute of Technology, Sweden
Inspired by the fast evolution of the emerging digital technologies such as Internet of Things, 5G, cloud computing, and artificial intelligence, the manufacturing industry are calling for new generation industrial automation systems (IAS) that can be deployed on open, flexible, and IT-style communication and computing infrastructures. To fulfill this demand, we have proposed the new paradigm of Cloud Fog Automation, in which the higher-level software functionalities of IAS (i.e., Level 3&4 in the ISA-95 pyramid) are hosted in general purpose cloud computing infrastructure, and more importantly, the lower-level time-critical control functionalities (i.e., Level 1&2 in the ISA-95 pyramid) are also hosted by virtualized computing infrastructure which we call “fog” in contrast with “cloud”.
However, major technical challenges in the three “C” subjects, i.e., communication, computing, and control, must be solved before the expected benefits are achievable, especially in the Level 1&2 time-critical control that requires stringent determinism. In the past decades, these challenges had been addressed separately in the three subjects but without success, unfortunately. From now on, we believe the cross-disciplinary research, or so-called Communication-Computing-Control Co-Design (3C Co-Design), will be the most promising strategy to realize the Cloud Fog Automation.
In this presentation, I will share what we have done and the open challenges in this direction. I will show the latency and reliability of the latest wireless technologies such as 5G and WiFi6 and their insufficiencies, then experimental results of using open and virtualized computing platform to host time critical control applications. One step further, I will show significant improvements in the overall control performances by the 3C Co-Design, e.g., tuning the control model according to the latency pattern of the communication and computing infrastructure. Despite the encouraging progresses, more challenges are still open to generalize the 3C Co-Design approach. I hope to trigger more discussions on this topic by this talk.
Dr. Zhibo Pang received MBA from University of Turku in 2012 and PhD from KTH Royal Institute of Technology in 2013. He is currently a Senior Principal Scientist at ABB Corporate Research Sweden, and Adjunct Professor at KTH. He is a Member of IEEE IES Industry Activities Committee, Steering Committee Member of the IEEE IoT Technical Community, Vice-Chair of the TC on Cloud and Wireless Systems for Industrial Applications, and Co-Chair of the TC on Industrial Informatics. He is Associate Editor of IEEE TII, IEEE JBHI, and IEEE JESTIE. He was Adjunct Professor at University of Sydney from 2019 to 2023. He was General Chair of IEEE ES2017, General Co-Chair of IEEE WFCS2021, and Invited Speaker at the Gordon Research Conference AHI2018. He was awarded the “Inventor of the Year Award” by ABB Corporate Research Sweden, three times in 2016, 2018, and 2021 respectively. He works on enabling technologies in electronics, communication, computing, control, artificial intelligence, and robotics for Industry4.0 and Healthcare4.0. He has 24 granted patents, 100+ refereed journal papers and 50+ refereed conference papers.