Title: Humanoids as General-Purpose Robots 

Abstract: 

Humanoid robots (humanoids) are rapidly gaining more maturity in terms of hardware and embedded software. Recently, we have witnessed substantial improvement in existing series and the revelation of new platforms. So far, the development of humanoids has been confined to research, and there is still much to do in several specific aspects in order to engage in an innovation pathway toward fully industrial or domestic products. The first revealed applications and business cases of humanoid robots are oriented towards entertainment and disaster operations. In previous years, we focused our efforts on applications of humanoids in manufacturing and domestic assistance for frail persons. My talk will summarize the achievements towards that purpose and discuss perspectives in science, implying advances in machine intelligence.

Short biography:

Professor Abderrahmane Kheddar received the B.S. degree in computer science from the Institut National d’Informatique, Algiers, Algeria, in 1990, and the M.Sc. and Ph.D. degrees in robotics from Pierre et Marie Curie University, Sorbonne University, Paris, France in 1993 and 1997, respectively. In 2008, he created the CNRS-AIST Joint Robotic Laboratory, an International Research Laboratory, located in Tsukuba, Japan, where he was the Director from 2008 to 2016 and Codirector from 2017 to 2021. In 2010 he also created and led the Interactive Digital Humans team until 2020, with the Laboratory of Computer Science, Robotics and Microelectronics of Montpellier, CNRS, University of Montpellier, France. His research interests include haptics, humanoids, and related bionics. He is a Founding Member of the IEEE Robotics and Automation Society (RAS) Chapter on Haptics, and the Co-Chair and Founding Member of the IEEE RAS Technical Committee on Model-Based Optimization. He is a Member of the Steering Committee of the IEEE Brain Initiative, an Editor of the IEEE Robotics and Automation Letters, and a Founding Member and the Deputy Editor-in-Chief for Cyborg and Bionics System. He was an Editor of the IEEE Transactions on Robotics, from 2013 to 2018. He is a Founding Member of the IEEE Transactions on Haptics and was in its Editorial Board from 2007 to 2010. Since 2020 he is the lead of the bionics initiative at CARTIGEN, University Hospital of Montpellier. He is a Fellow of the IEEE, a Fellow of the Asia-Pacific Artificial Intelligence Association and Vice-President of the International Artificial Intelligence Industry Alliance (AIIA). He is a Full Member of the National Academy of Technology of France and a Knight of the National Order of Merits of France.

Title: Recent Advances in Humanoid and Legged Locomotion: From Model-based Optimization to Data-driven Reinforcement Learning

Abstract:

Locomotion is among the most important skills for humanoid and legged robots. Such a skill enables the robots to reliably move in complex environments and allows for efficient integration with upper layer manipulation skills to achieve general mobile manipulation tasks. Thanks to the recent advancements in both hardware systems and algorithmic design, locomotion ability for humanoids has been significantly improved. From the methodological perspective, we are witnessing a clear trend of moving from model-based optimization and model predictive control to data-driven reinforcement learning. In this talk, I will briefly talk about our recent works on the development of reliable and robust locomotion controllers for humanoid and bipedal robots, leveraging both optimization-based and reinforcement learning techniques.

Short biography:

Hua Chen receives the B.E. degree in Automation from the College of Control Science and Engineering in Zhejiang University, Hangzhou, China in 2012, and the Ph.D. degree in Electrical and Computer Engineering from The Ohio State University, Columbus, Ohio in 2018. He was a post-doctoral researcher with the same University from Jan. 2019 to June. 2019. Then, he joined Southern University of Science and Technology as a research assistant/associate professor. He is currently an Assistant Professor with the Zhejiang University-University of Illinois Urbana-Champaign Institute and is also a co-founder of LimX Dynamics. His research interests lie in the interdisciplinary area between control theory, artificial intelligence, and robotics, with an emphasis on humanoids. He is a member of IEEE and serves as the associate editor for IROS, ICCA, and reviewer for various journals and conferences.

Title: Momentum Control of Humanoid Robots: Walking, Jumping, and Disturbance Rejection

Abstract:

In recent years, humanoid robots have made explosive advancements in their mobility capabilities, including object grasping control, walking control, jumping control, disturbance rejection control, and motion adaptation in various terrains. These improvements are due to advancements in robot modeling and planning methods as well as enhancements in robotic hardware and control methods. A particularly noteworthy aspect is the application of whole-body momentum in robot control. Compared to quadruped robots, biped robots have limbs with greater mass and more unstable dynamics, which necessitates considering more nonlinear system modeling and more complex balancing strategies in their control. My talk will introduce our practical work in disturbance resistance simulation, single-leg robot jumping control, and humanoid robot jumping control, highlighting the unique advantages of incorporating momentum into full-body control of robots.

Short biography: 

Dr. Zhicheng He graduated from the School of Computing at Harbin Institute of Technology and currently serves as the Director of Algorithms at Leju (Shenzhen) Robotics Co., Ltd. Throughout his career, he has applied for and obtained more than ten patents related to robotics inventions and has published several SCI-indexed papers. His projects have included the design and control of a flexible arc-legged hexapod robot system, the design of a hopping leg robot, as well as environmental perception and gait planning for bipedal robots. Under his leadership, the Leju Roban robotics research and development team has successfully implemented the application of small and medium-sized robots in complex environments, such as the ability to climb stairs and walk over plum blossom piles. Moreover, he has led the development of the KUAVO high-dynamic humanoid robot, which is capable of fast squatting, jumping, and robust walking, demonstrating a high degree of technical innovation and practicality. His current research interests focus on the design of lightweight and rigid-flexible coupled humanoid robots, the motion control of high-dynamic humanoid robots, and the integration of reinforcement learning, imitation learning, and optimal control methods.

Title: The perception for humanoid robots and its challenge

Abstract:

Deep learning has achieved remarkable success in various visual and language tasks; however, these approaches are typically limited to single tasks and modalities. As for the control of robots, it requires real-time and precise perception of the environment. The main challenge lies in two aspects: how to balance accuracy and computational complexity, and how to integrate the multimodal input and provide required information for the downstream tasks. In the past few years, we have focused on model compression for stereo matching and multi-task learning for autonomous driving.  This talk will introduce our methods and achievements, as well as the current state of the industry.

Short biography:

Dr. Baiyu Pan received the B.S. degree in computer science and information from Jilin University, in 2013, and the M.Sc. and Ph.D. degrees in computer science from University of Macau in 2016 and 2021, respectively. He joined the research institute of UBTech Corporation, where he was research engineer from 2021 to 2023, and team leader from 2024. He is now leading the team of multimodal perception. In 2022, he also started as a postdoc at Shenzhen Institute of Advanced Technology. His research interests include stereo matching, multimodal perception, and multimodal LLM.

Title: Learning fault-tolerant locomotion

Abstract:

Reinforcement learning is playing an increasingly important role in robotic control. Deep reinforcement learning enhances the stability of quadruped robots by training them in simulated environments and then transferring the learned policies to the real world. By using trained neural networks, quadruped robots are able to walk stably on various terrains. However, current reinforcement learning techniques primarily focus on improving the stability of walking under normal conditions for quadruped robots. This talk will explore the challenges of robot walking when their motors experience partial or total failures.

Short biography:

Professor Peng Lu obtained his BSc degree in automatic control and MSc degree in nonlinear flight control both from Northwestern Polytechnical University (NPU). He continued his journey on flight control at Delft University of Technology (TU Delft) where he received his PhD degree. After that, he shifted a bit from flight control and started to explore control for ground/construction robotics at ETH Zurich (ADRL lab) as a Postdoc researcher. He also had a short but nice journey at University of Zurich & ETH Zurich (RPG group) where he was working on vision-based control for UAVs as a Postdoc researcher. He was an assistant professor in autonomous UAVs and robotics at Hong Kong Polytechnic University prior to joining the University of Hong Kong in 2020. He has received several awards such as 3rd place in 2019 IROS autonomous drone racing competition. He serves as an associate editor for 2020 IROS and session chair/co-chair for conferences like IROS and AIAA GNC for several times. He also gave a number of invited/keynote speeches at multiple conferences, universities and research institutes.

Title: Wearable Assistive Robots for Aging Society

Abstract:

A rapidly aging population is one of the grand challenges facing the society. The number of people aged 65 or older worldwide is estimated to reach 1.6 billion by 2050. A major difficulty that many older people experience is severe limitation in mobility and manipulability in their daily life, resulting in tremendous social and economic challenges. This talk will discuss a User-Centric Co-Creation (UC³) approach to develop intelligent robotic systems to assist mobility and manipulability and prevent falls. The UC³ methodology lays down a theoretical foundation for multi-disciplinary approach to develop personalized wearable assistive systems. It will pave a new avenue to advance ergonomics and gerontechnology beyond current horizons.

Short biography:

Dr. Jiang-Cheng Chen received his Ph.D. degree in mechanical engineering from Xi'an Jiaotong University, Shaanxi, China, in 2016. He served as a postdoctoral fellow in the Department of Industrial and Manufacturing System at The University of Hong Kong, Hong Kong, from April 2017 to April 2018, and in Shenzhen Academy of Robotics, China, from May 2018 to July 2021. From August 2021 to November 2022, he held the position of tenure-track assistant professor at the College of Mechatronics and Control Engineering at Shenzhen University, Shenzhen, Guangdong, China. Currently, he is a research assistant professor at the Department of Industrial and Manufacturing System, The University of Hong Kong, Hong Kong, HKSAR, China.