20点直播｜香港城市大学、北京大学等三位专家报告

直播时间：2024年4月30日（周二）20:00-22:00

直播平台：

科学网APP

（科学网微博直播间链接）

科学网微博

科学网视频号

北京时间4月30日晚八点，iCANX Youth Talks第五十三期邀请到了香港城市大学Shigi Wang，北京大学Chang Liu，北京邮电大学Bin Fang三位教授主讲，清华大学Xiaonan Wang教授作为主持人，北京大学Haixia Zhang教授担任嘉宾，期待你一起加入这场知识盛宴。

【嘉宾介绍】

Shiqi Wang

香港城市大学

Generative Visual Compression: Promises and Challenges

【Abstract】

Recent years have witnessed the rapid development of generative AI. In this talk, I will present our research activities for generative visual data compression, in particular face compression relying on strong prior knowledge. More specifically, we will first discuss the ultra-low bitrate face video compression with compact temporal trajectory representation. Subsequently, we show how to achieve interactive face video compression with compact and controllable facial semantics. Finally, we discuss the quality assessment of face generative coding and related standardization activities. We will further share our thoughts on the future development of generative visual compression.

近年来，生成式人工智能技术发展迅速，为视频编码和表示带来了新的机遇和挑战。其中具有强先验信息的人脸压缩更是取得了较VVC等更为优异的率失真性能。本次报告将分享我们在基于人工智能的生成式压缩编码及质量评价方面的相关研究成果和思考，尤其以人脸为例，探讨具有可扩展编码能力和面向视觉分析的表征算法、基于紧凑特征表达和面向超低码率的视频压缩方案、支持语义可控和面向交互的视频压缩算法、面向元宇宙的3D语义表征及分析方案、面向生成式压缩的视觉质量评价指标以及生成式压缩的标准化工作。

【BIOGRAPHY】

Shiqi Wang received the Ph.D. degree in computer application technology from Peking University in 2014. He is currently an Associate Professor with the Department of Computer Science, City University of Hong Kong. He has proposed more than 70 technical proposals to ISO/MPEG, ITU-T, and AVS standards. He authored or coauthored more than 300 refereed journal articles/conference papers, including more than 100 IEEE Transactions. His research interests include video compression, image/video quality assessment, video coding for machine, and semantic communication. He received the Best Paper Award from IEEE VCIP 2019, ICME 2019, IEEE Multimedia 2018, and PCM 2017. His coauthored article received the Best Student Paper Award in the IEEE ICIP 2018. He served or serves as an Associate Editor for IEEE Transactions on Circuits and Systems for Video Technology, IEEE Transactions on Image Processing, IEEE Transactions on Cybernetics, IEEE Transactions on Multimedia, IEEE Access, APSIPA Transactions on Signal and Information Processing.

王诗淇，2014年毕业于北京大学，获得博士学位。目前任香港城市大学副教授，博士生导师。从事视频编码与处理、人工智能、计算机视觉等方面的研究。在国内外期刊和会议上发表/接收论文300余篇，其中包括IEEE 汇刊论文100余篇。论文谷歌引用13000余次。目前担任人工智能、视频处理领域旗舰期刊IEEE-TIP, TMM, TCyber和TCSVT 编委，并获得IEEE ICME, VCIP, MultiMedia等多个国际会议和期刊的最佳论文奖。王诗淇博士的研究也获得香港研究资助局杰出青年学者计划等多项项目资助。

Chang Liu

北京大学

Robot collaborative motion planning and active sensing

【ABSTRACT】

The rapid development of artificial intelligence technology has not only driven continuous innovation in the field of intelligent robotics but has also led to widespread applications of robotics technology in various areas such as intelligent transportation, high-end manufacturing, and smart healthcare, creating profound social impacts and significant economic value. Perception and motion planning are the two core functionalities of intelligent robots, mutually driving and complementing each other, forming the foundation of intelligent robotic behavior. On one hand, robust perception and information fusion capabilities provide crucial environmental information for motion planning, which is essential for ensuring robots navigate safely and efficiently in complex environments. On the other hand, well-designed motion trajectories not only enhance the efficiency of robot movements but also strengthen the environmental information obtained through perception, thus requiring optimization of motion trajectories to improve robots perception of their surroundings. This talk will focus on the speakers research achievements in these two areas. Firstly, addressing the challenges of motion planning faced by robot swarms in unknown environments, we propose an innovative and efficient online motion planning method. This method enables robot swarms to adaptively plan their movements to cope with changing environmental conditions. Then, we will use the target identification and search, a critical topic in the field of active sensing, as an example to introduce a navigation path planning scheme driven by foundation models. This scheme effectively integrates environmental semantic information and common-sense reasoning capabilities to improve the efficiency of search tasks.

人工智能技术的飞速发展，极大推动了智能机器人领域的持续创新，使得机器人技术在智能交通、高端制造、智慧医疗等诸多领域得到广泛应用，产生了深远的社会影响，创造了显著的经济价值。感知和运动规划是智能机器人两个核心功能，两者相互牵引、相辅相成，共同构成了机器人智能行为的基础。一方面，良好的感知和信息融合能力，能为运动规划提供关键环境信息，这些数据是确保机器人安全、高效地在复杂环境中导航的前提。另一方面，合理设计的运动轨迹不仅能提升机器人行动效率，还能增强其感知所获取的环境信息，因而需要通过优化运动轨迹来提高机器人对环境的感知效果。本次报告将重点介绍报告人在上述两方面的研究成果。首先，针对机器人集群在未知环境中面临的运动规划挑战，我们提出一种创新的高效在线运动规划方法，能够使机器人集群实现自适应的运动规划，以应对不断变化的环境条件。然后，我们将以目标识别与搜索这一主动感知领域的重要课题为例，介绍一种基于大模型驱动的导航路径规划方案，能有效融合环境语义信息和常识推理能力，提高搜索任务效率。

【BIOGRAPHY】

Chang Liu is an Assistant Professor and doctoral supervisor in the College of Engineering at Peking University, China. He is a member of the national youth project of the overseas talents program and of the Beijing Nova program. He received the B.S. degrees in Electronic Information Science and Technology and in Mathematics and Applied Mathematics (double degree) from the Peking University, China, and the M.S. degrees in Mechanical Engineering and in Computer Science, and the Ph.D. degree in Mechanical Engineering from the University of California, Berkeley, USA. He was a Postdoctoral Associate in the Sibley School of Mechanical and Aerospace Engineering at Cornell University. He has worked at Ford Motor Company and NVIDIA Corporation on developing autonomous driving systems. His research focuses on robot motion planning, optimal filtering, active sensing, and multi-robot coordination. His contributions have been recognized by several academic awards including the best paper award from IEEE T-IV, ICCAS best student paper award, and ISRR Blue Sky Ideas award. He has served as an area chair at RSS 2024, associate editor of both IROS 2023 and IROS 2024, and a member of program committee at MRS 2019.

刘畅，北京大学工学院助理教授、博士生导师，北京市科技新星。本科毕业于北京大学，获电子信息科学与技术学士学位以及数学与应用数学双学士学位。博士毕业于美国加州大学伯克利分校，获机械工程专业博士学位。读博期间获得该校机械工程与计算机科学两个硕士学位。曾在美国康奈尔大学机械与航空工程系任博士后，并曾工作于美国福特汽车公司和美国英伟达公司，承担自动驾驶系统的研发任务。刘畅目前的研究主要集中在机器人运动规划、最优滤波、主动感知和多机协同等方面，获得包括IEEE T-IV最佳论文奖，ICCAS最佳学生论文奖，以及ISRR“蓝天创意”论文奖等在内的多项奖励。担任国际机器人会议RSS 2024领域主席，IROS 2023和2024副编委，MRS 2019程序委员会委员。

Bin Fang

北京邮电大学

Embodied Tactile and RobotGPT for Manipulation

【ABSTRACT】

Intelligent Dexterous Manipulation is rated by Science as one of the most challenging problems in the field of robotics, and is the core key technology of the national major strategy of intelligent robotics and artificial intelligence. It is also listed as an important development direction by OpenAI, Google, NVIDIA and other cutting-edge technology companies in the world. As one of the important key supporting technologies, embodied tactile is the bridge connecting the external environment and the robot body. "Flexible Smart Tactile Sensor" has also been selected as one of the 2023 Global Engineering Frontier Development Frontier Directions. This report focuses on the principles and methods of embodied intelligent taptic sensing, introduces the basic principles of typical haptic sensing, elaborates on the methods of visual based tactile sensing, and demonstrates a series of tactile sensing devices and related research work on the virtual-real migration of manipulation skills. It further outlines the recent research progress in large model of robot manipulation, introduces the research on RobotGPT for robot manipulation learning framework, and discusses the development trend of future technologies.

具身智能灵巧操作被Science评为机器人领域最具挑战问题之一，是智能机器人与人工智能国家重大战略的核心关键技术。也被OpenAI、谷歌、英伟达等世界前沿科技公司列为重要发展方向。具身触觉作为重要的关键支撑技术之一，是连接外部环境与机器人本体的桥梁。“柔性智能触觉传感器”也入选了2023全球工程前沿领域开发前沿方向。本报告主要围绕具身智能触觉感知原理和方法展开，介绍了典型触觉传感的基本原理，详细阐述视触觉感知方法，展示了系列视触觉传感装置和操作技能虚实迁移的相关研究工作。进一步概述近期机器人操作大模型的研究进展，介绍了面向机器人操作学习框架的RobotGPT研究情况，并探讨未来技术的发展趋势。

【BIOGRAPHY】

Bin Fang is a professor at Beijing University of Posts and Telecommunications. His main research interests include intelligent perception, interaction and operation of robots. He is also the Secretary General of the Cognitive Systems and Information Processing Committee of the Chinese Society for Artificial Intelligence, a distinguished member of the Chinese Society for Artificial Intelligence, an executive member of the Intelligent Robotics Committee of the Chinese Computer Society, and a senior member of the IEEE. He has published hundreds of papers in high-level journals such as Nature Communications, IEEE TRO, Soft Robotics, and conferences such as ICRA, and has published one monograph in English and one in Chinese. He has received 8 best paper awards from international conferences/journals, including ICRA2021 Best Paper Nomination Award at the Robotics Summit, ICARM2021 Best Paper Award, ROBIO2012 Best Paper Award, ICARM2012 Best Paper Award, and ICARM2012 Best Paper Award. Paper Award, ROBIO2019 Best Student Paper Award, etc. Organised special issues on Robot Dexterous Manipulation in several SCI international journals as chief guest editor-in-chief. He has led his students to win several championships in international IROS grasping competitions, and was awarded "Excellent Instructor" by Tsinghua University. His tactile dexterous hand was selected for the National 13th Five-Year Plan Science and Technology Innovation Achievement Exhibition, and his autonomous pharyngeal swab collection robot system won the silver medal at the International Geneva Invention Exhibition. At the same time, he won the "Early Career Award" of IEEE Bionic Electromechanics and Robotics Technical Committee, "Young Scientist Award" of China Command and Control Society, "Innovative Craftsman" of Beijing Municipality, "Natural Craftsman" of Chinese Society of Automation, "Innovative Craftsman" of Beijing Municipality and "Innovative Craftsman" of Beijing Municipality. "He also won the first prize of Natural Science Award of Chinese Society of Automation.

方斌，北京邮电大学“拔尖人才”教授，主要研究方向机器人智能感知、交互与操作。兼任中国人工智能学会认知系统与信息处理专委会秘书长，中国人工智能学会杰出会员, 中国计算机学会智能机器人专委会常务委员，IEEE 高级会员。发表Nature Communications、IEEE TRO、Soft Robotics等高水平期刊及ICRA等会议论文上百篇，出版中英文专著各1部，获得国际会议/国际期刊最佳论文奖8项，包括机器人领域顶会ICRA2021最佳论文提名奖、ICARM2021最佳论文奖、ROBIO2019最佳学生论文奖等。作为首席客座主编在多个SCI国际期刊组织了机器人灵巧操作的专刊。多次带领学生在国际IROS抓取操作比赛中获得冠军，曾被清华大学评为“优秀指导教师”。研制的触感灵巧手入选了国家“十三五”科技创新成就展，开发的咽拭子自主采集机器人系统获得了国际日内瓦发明展银奖。同时获得了IEEE仿生机电及机器人技术委员会“早期职业生涯奖”、中国指挥控制学会“青年科学家奖”、北京市“创新大工匠”，中国自动化学会自然科学奖一等奖等。

【主持人】

Xiaonan Wang

清华大学

【研讨嘉宾】

Haixia Zhang