Plenary/ Invited Speakers
Dongsoo Har, KAIST, Korea
Biography: Prof. Dongsoo Har received the B.Sc. and M.Sc. degrees in electronics engineering from Seoul National University, respectively, and the Ph.D. degree in electrical engineering from Polytechnic University, Brooklyn, NY, USA. He has worked for AirTouch, Vodafone(US), and GIST. He is now the Director of Center of Green Transportation System Development in KAIST. He was a recipient of the Best Paper Award (Jack Neubauer Award) from the IEEE Transactions on Vehicular Technology in 2000. He is an associate editor of IEEE Sensors Journal. He authored and co-authored more than 100 articles published in international journals and conferences. He was the member of advisory board, program chair, vice-chair, and general chair of international conferences. He served as a vice-chair of AI World Cup in 2018. His research interests include theory and applications of artificial intelligence.
AI Robot Soccer : Multi-agent Cooperative Game
John C. Gallagher, Wright State University in Dayton, Ohio USA
Biography: Dr. John C. Gallagher is a professor in the Department of Computer Science and Engineering at Wright State University in Dayton, Ohio USA. He received his Ph.D. in computer engineering from Case Western Reserve University in Cleveland, Ohio and for the past twenty years has investigated machine learning as applied to walking and flapping-wing flying robots, analog neuromorphic computation with both software and hardware implementations, and metaheuristic search for dynamic optimization problems. He is a senior member of the IEEE and has maintained twenty years of uninterrupted support for various projects from the US National Science foundation, including an NSF CAREER award. He has also been active in a number of projects with the US Air Force and regularly consults on practical problems in machine learning, sensing, and control of electromechanical systems.
Integrated Model Inference and Checking for Long Term Safety Assurance of Flapping-Wing Vehicles Containing Locally Adapting Smart Components
Abstract: Roboticists have long argued for the use of in-service adaptation of robotic systems as a means of restoring correct operation in the face of ongoing damage or unexpected environmental changes. Such adaptation is often implemented via the use of “smart components” that modify local structure to optimize some measure global system quality. This practice, as effective as it might be, may introduce unmodeled interactions into the system and prevent the ongoing use of otherwise applicable Verification and Validation (V&V) techniques. Without such use of such techniques, it could be difficult to ensure the long-term safety and efficacy of temporally and organizationally local modifications when viewed from the global perspective of long-term robot behavior. In this talk, we will examine methods to alleviate V&V difficulties by adopting a holistic approach in which we melt the traditional boundaries among system adaption, system identification, and model checking. We will explore the methods in the context of integrated adaptation, model extraction, and model checking for an operational flapping-wing flying vehicle flying typical mission profiles. We will further discuss how similar boundary dissolving could be of use for providing safe and effective in-service adaption for other robotic systems.
Dongjun Lee, Seoul National University, Korea
Biography: Dongjun Lee (B.S. KAIST, 1995, M.S. KAIST, 1997, Ph.D. Minnesota 2004) is a Professor with the Department of Mechanical & Aerospace Engineering at the Seoul National University (SNU). Prior to that, he was an Assistant Professor at the University of Tennessee – Knoxville from 2006 to 2011, a Postdoctoral Fellow with the Coordinated Science Laboratory at the UIUC from 2004 to 2006, and an Engine Development Engineer at Kia Motors, Korea, from 1997 to 1999. His main research interests include dynamics and control of robotic and mechatronic systems with particular emphasis on aerial/mobile robots, teleoperation/haptics, multirobot systems, and industrial control applications. Dr. Lee is a recipient of US NSF CAREER Award 2009, Best Paper Award from IAS 2012, Best Video Award from URAI 2013, and Doctoral Dissertation Fellowship of the University of Minnesota 2002. He was an Associate Editor of IEEE Trans. on Robotics 2010-2014, an Area Chair of RSS 2015, an Editor of ICRA 2015-2017 and an Associate Editor of IEEE Trans. on Haptics.
Design, Estimation and Control of a New Aerial Platform, LASDRA (Large-size Aerial Skeleton with Distributed Rotor Actuation)
Abstract: A dream of many roboticists is to realize megabots (e.g., 15m, 20-DOF, able to fly), yet, real-world robots typically are too heavy with only limited-DOF (e.g., KUKA L750 Titan (3.6m, 8-DOF, 5T), Method 2 (4.1m, 24-DOF, 1.6T), etc.). This discrepancy, we believe, is due to the fundamental limitation of ubiquitous electrical motor and hydraulic actuations, namely, with both being “internal actuation”, all self-weight and external force are accumulated to their base, thereby, requiring excessively bulky/heavy construction. In this talk, I will introduce a new aerial platform. LASDRA, which, by utilizing distributed rotors as “external actuation”, can realize very long/slender, yet, still light/dexterous robotic systems. Implementations of this LASDRA for long-reach horizontal compliant manipulation/operation and for dragon-like articulated autonomous flying will be presented with the relevant design, estimation and control challenges elucidated. Some other results of my laboratory related to those will also be introduced.
Jun Jo, School of ICT, Griffith University, Australia
Biography: Assoc. Professor Jun Jo was awarded his PhD degree from the University of Sydney. He has been working on computer vision and machine learning and their applications in various areas including satellite image analysis and MRI image analysis. He has published over 150 refereed publications. Dr Jo is currently taking the position of the President of Australian Robotics Association (ARA). Dr Jo is also the Program Director for Bachelor of Intelligent Digital Technologies (BIDT) degree program at Griffith University, Australia.
Bushfire Surveillance from the Space
Abstract: Bushfires, or wildfires, are frequent events especially during the hot and dry months of the year. Such fires impact extensive areas, causing severe property damage and loss of human life. Due to their fast-spreading nature, wildfires are often detected when already beyond control and consequently cause billion-scale effects in a very short time. Governments are looking for remote sensing methods for early wildfire detection, avoiding billion-dollar losses of damaged properties.
A research team at Griffith University is developing an autonomous and intelligent system built on top of imagery data streams, which is available from around-the-clock satellites, to monitor and prevent fire hazards from becoming disasters. However, satellite data pose unique challenges for image processing techniques, including temporal dependencies across time steps, the complexity of spectral channels, and adversarial conditions such as cloud and illumination. In this talk, I will discuss a novel multi-modal learning method that combines both satellite images and weather data in an advanced deep learning architecture for detecting and locating wildfires at both image and pixel level. Our system is built and tested on the Geostationary Operational Environmental Satellites (GOES-16) and Himawari-8 streaming data sources. Some experimental results will be presented and discussed.
Anwar P.P. Abdul Majeed, Universiti Malaysia Pahang, Malaysia
Biography: Dr. Anwar P.P. Abdul Majeed graduated with a first class honours B.Eng. in Mechanical Engineering from Universiti Teknologi MARA (UiTM), Malaysia. He obtained an MSc. in Nuclear Engineering from Imperial College London, United Kingdom. He then received his PhD in Rehabilitation Robotics under the supervision of Prof. Dr. Zahari Taha from Universiti Malaysia Pahang (UMP). He is currently serving as a senior lecturer at the Faculty of Manufacturing Engineering, UMP. He is an active research member at the Innovative Manufacturing, Mechatronics and Sports Laboratory, UMP. His research interest includes computational mechanics, applied mechanics, sports engineering, renewable and nuclear energy, sports performance analysis as well as machine learning.
Machine Learning in Sports
Abstract: The application of machine learning in sports is non-trivial as the nature of the data involved in sporting activities are more often than not, complex, and hence, the conventional statistical techniques often fall short in providing meaningful insights. This talk will elucidate the application of various machine learning algorithms in evaluating sporting activities both in individual and team sports. This talk will be beneficial to relevant stakeholders embracing such state of the art evaluation techniques.
Jaesik Choi, KAIST, Korea
Education- University of Illinois at Urbana-Champaign, Ph.D. in Computer Science, May 2012 – Seoul National University, B.S. in Computer Engineering, magna cum laude, August 2004.
Bio- Director, UNIST Explainable Artificial Intelligence Center, established by Ministry of Science and ICT, July 2017 – present.- Rising Star Distinguished Professor, UNIST, September 2018 – present.- Associate Professor, School of Electrical and Computer Engineering, UNIST, September 2017 – present.- Assistant Professor, School of Electrical and Computer Engineering, UNIST, July 2013 – August 2017.- Research Affiliate, Lawrence Berkeley National Laboratory, July 2013 – October 2018.- Postdoctoral Fellow, Lawrence Berkeley National Laboratory, January 2013 – July 2013.
Explainable Artificial Intelligence for Reinforcement Learning
Abstract: As many complex AI systems are deloveped and used in our daily lives. It becomes important to interpret and explain the decision of complex AI systems. In this talk, I will overview various methods and perspective in explainable Artificial Intelligence (XAI) methods. Topics include interpretable deep neural networks models, Bayesian model compositions, and model agnostic approaches. Then, I will introduce recent methods to interpret complex decision in deep reinforcement learning.