PCC Young Scholars Forum—A New Era of Urban Systems Driven by Ubiquitous Intelligence
August 22 (Friday) 1:00 PM-3:00 PM
Location: Shanghai Hall, 3rd Floor

Introduction:
        Assistant Professor at the Center for Intelligent Transportation Systems, Hong Kong University of Science and Technology, Guangzhou. His research interests focus on the interdisciplinary fields of intelligent IoT, data science, and intelligent transportation. His research has been published in leading international journals and conferences, including IEEE TMC, IEEE TKDE, IEEE ToN, IEEE TITS, IEEE TBD, ACM KDD, ACM MobiSys, IEEE ICDE, ACM MobiHoc, and IEEE ICDM. His research has won the "Best Paper Award" at IEEE ICPADS '20. He has been invited to serve on the program committees of leading international conferences, including ACM KDD, ACM WSDM, and IEEE ICDCS, and has been a long-time reviewer for important international journals, including IEEE TKDE, IEEE TMC, and IEEE TITS. He has led eight projects funded by the National Natural Science Foundation of China and at provincial and municipal levels.

Report Title: Research on AI-Driven Ridesharing

Report Introduction: 

        As global carbon neutrality goals and intelligent transportation systems deepen, ridesharing, as a core smart transportation scenario, leverages the deep integration of ubiquitous computing, the Internet of Things, and 5G/6G communication technologies to become a key path to alleviating urban traffic congestion and promoting low-carbon transformation. However, in dynamic and complex scenarios, insufficient positioning accuracy due to GNSS signal attenuation in urban canyons, inefficient matching caused by the spatiotemporal heterogeneity of driver and passenger demand, and scheduling complexity caused by order-vehicle coupling continue to severely constrain system operational efficiency and service experience. This report focuses on the innovative application of AI technology across the entire shared mobility process, specifically reporting on research progress in three areas: 1) precise positioning in urban canyons driven by multi-source data fusion; 2) efficient driver-passenger pre-matching using spatiotemporal attention mechanisms; and 3) joint optimization strategies for order dispatch and vehicle scheduling to improve system efficiency and user experience. Furthermore, the report explores the potential of large-scale models in shared mobility scenarios and the key challenges they face.

Introduction:
        Dr. Liu Hao is currently an Assistant Professor, Doctoral Supervisor, and Director of Graduate Studies in Artificial Intelligence at the Hong Kong University of Science and Technology, Guangzhou. He was selected for the National High-level Young Talent Program and the Forbes China 30U30 list. He was formerly a Senior Researcher at Baidu Research Institute. His research interests focus on spatiotemporal data mining and urban infrastructure models. In the past five years, he has published over 80 papers in leading AI and data mining conferences and journals, including SIGKDD, VLDB, NeurIPS, ICML, and TKDE. He has been granted over 40 patents in China and the US, and has been nominated for the VLDB Best Paper Award, as well as the Special Award and Gold Medal at the 50th Geneva Invention Exhibition. He serves as Associate Editor of IEEE Transactions on Big Data, an Editorial Board Member of npj | Artificial Intelligence, a Young Editorial Board Member of JCST, the IJCAI 2025 Competition Chair, and the VLDB 2024 Local Chair. Many of his AI technologies have been successfully implemented in internet, social, and government projects at companies like Didi Chuxing, Baidu, and Xinhua News Agency, impacting hundreds of millions of Chinese citizens.

Report Title: Constructing and Applying Urban Intelligent Agents

Report Introduction: 

       Continuous breakthroughs in large models are reshaping urban intelligent systems. Current spatiotemporal data mining and machine learning methods face a significant theory-practice gap in open urban scenarios, severely restricting their practical application. This report will introduce the core challenges facing urban intelligent systems in open urban environments, explore different paradigms for building urban intelligent agents based on large language models, and report on the team's recently proposed application of intelligent agents for urban knowledge understanding and decision optimization.

Introduction:
       Postdoctoral Fellow in the Department of Electronic Engineering at Tsinghua University. He received his bachelor's and doctoral degrees from the Department of Electronic Engineering at Tsinghua University in 2016 and 2021, respectively. His research focuses on cutting-edge areas such as spatial intelligence, urban science, spatiotemporal data mining, large language models, and intelligent agents. He has published over 40 high-quality papers in conferences and journals, including KDD, ICCV, ACL, NAACL, WWW, AAAI, and TKDE. His papers have received over 3,800 citations on Google Scholar, with nine of his papers exceeding 100 citations. His DeepMove method (WWW 2018) has become a classic algorithm for modeling human mobility behavior, receiving over 800 citations. He was selected as one of the 2024 Stanford Top 2% Global Scientists and received support from the 2024 Tsinghua University Shuimu Scholar Program and the 2024 National Postdoctoral Researcher Program. He is also a member of the ACM SIGSPATIAL China Executive Committee and was named an ACM SIGSPATIAL China Spatial Data Intelligence Rising Star Scholar.

Report Title: Large-Scale Agent-Driven Individual Mobility Behavior Modeling in Urban Environments

Report Introduction:

       Human mobility behavior modeling is a core research topic in spatial data intelligence. Over the past decade, deep learning-based modeling paradigms have made significant progress. However, this approach still faces numerous challenges, such as the scarcity of high-quality training data, the difficulty of modeling the complex underlying behavioral mechanisms of individuals, and insufficient cross-regional spatial generalization. In recent years, breakthroughs have been made in large-scale models and agent-based technologies, demonstrating significant potential in world knowledge, generalization, and reasoning and planning. These technological advances provide new opportunities to address bottlenecks in mobility behavior modeling. This report explores a new paradigm for modeling individual mobility behavior using large models and intelligent agent technology. It explores its potential in modeling the underlying mechanisms of individual behavior, effectively addressing key challenges such as data loss and cross-regional generalization, particularly in the areas of spatiotemporal memory design and the utilization of world knowledge.

Introduction:
        Liu Jia is an Associate Professor and doctoral supervisor at Nanjing University. His research focuses on RFID passive sensing technology. In recent years, he has published over 80 academic papers in international conferences and journals such as NSDI, MOBICOM, MOBISYS, ATC, and SIGMOD, including over 40 CCF Class A papers. He holds over 20 authorized invention patents and three U.S. patents. This research has been recognized with awards such as the World's Top Ten Technological Advances in Intelligent Manufacturing, the Special Gold Medal at the Geneva International Invention Exhibition, the Jiangsu Province 333 Leading Talent Team, and the Huawei Spark Award. The research team independently developed an IoT mobile positioning system, which has been deployed in over 20 provinces nationwide and internationally, and participated in the drafting of a national standard that has been approved.

Report Title: Unveiling the Mystery of RFID's RSSI Compression

Report Introduction:

        In 2024, the number of newly added RFID tags worldwide exceeded 50 billion, making it the most widely used sensing technology in the IoT. RSSI, a key metric in wireless communication systems, directly impacts RFID sensing accuracy. This report, for the first time, reveals the "RSSI compression" effect in RFID backscatter communications. Due to the dynamic characteristics of the tag's radar cross-section (RCS), signal strength exhibits nonlinear systematic deviations, severely limiting sensing accuracy. The report will deeply analyze the physical mechanism of this phenomenon and propose a correction model based on dynamic RCS and an innovative single-sampling cutoff power measurement scheme, providing a strong guarantee for improving the perception accuracy of backscatter systems such as RFID.

Introduction:
        Guo Xiuzhen is a Hundred Talents Program researcher and doctoral supervisor at the School of Control Science and Engineering, Zhejiang University. Her research focuses on the Internet of Things, passive communications, and intelligent sensing. He has published over 50 papers in international conferences and journals, including NSDI, MobiCom, and TON. He has received awards such as the ACM MobiCom Best Paper Award, the IEEE HPCC Best Paper Award, and the IEEE MSN Best Student Paper Award. His research has been selected for the World of Science Conference (WOS) Highly Cited Papers and the ACM SIGMOBILE Research Hotspots. He was selected for the China Association for Science and Technology Young Talent Support Program and the Tsinghua University Shuimu Scholar Program. He received the Second Prize in Natural Sciences from the China Computer Federation and the 2021 ACM China Outstanding Doctoral Dissertation Award. He serves on the program committees of international academic conferences, including ACM MobiSys, ACM SenSys, IEEE ICDCS, and IEEE SECON.

Report Title: Research on Passive IoT Technologies Based on RF Computing

Report Introduction:

        In the era of the Internet of Everything, the limited sensing capabilities and high energy consumption of IoT technologies hinder their wider deployment and application. To address these challenges, we explore the dynamics and trends of IoT development from a computational perspective, exploring new IoT-focused computing theories and key technologies. This report will share our recent research findings in passive sensing networks and explore research directions in passive IoT technologies based on RF computing.

Introduction:
        Zhang Huanhuan is a distinguished researcher at the School of Computer Science and Technology, Beijing University of Posts and Telecommunications (a national demonstration software school). His research focuses on video IoT, intelligent network transmission, and system operations and maintenance. He has published over 20 papers (over 10 as first author) in top conferences and journals such as MobiCom, UbiComp, ToN, and TMC. He has applied for over 10 patents and participated in the development of one international ITU-T standard. He has received numerous honors, including the CCF Outstanding Doctoral Dissertation Incentive Program, the Ministry of Human Resources and Social Security Postdoctoral Innovation Talent Support Program, and the Second Prize for Scientific and Technological Progress of the China Electronics Society. He currently leads projects such as the National Natural Science Foundation of China Young Scientists Fund and Xiaomi's University-Enterprise Cooperation Program.

Report Title: Challenges and Optimization Methods for High-Quality Video Streaming in Mobile Interactive Scenarios

Report Introduction:

        In recent years, with the rapid development of mobile interactive applications such as HD video and cloud gaming, users have placed higher demands on low latency and a high-quality experience in video streaming. However, existing transmission mechanisms face significant bottlenecks in coping with network congestion and packet loss recovery, resulting in frequent video blurring and recovery times of several seconds, while also experiencing multiple freezes per minute. To address these challenges, our team collaborated closely with leading video transmission platforms. We conducted systematic measurements and analysis of interactive latency, image quality, and video frame characteristics under real-world network conditions. We proposed an intelligent video flow control algorithm (ACM MobiCom, ToN2024) and a refined FEC algorithm driven by frame-level data loss (NSDI 2025), effectively improving overall video quality and smoothness. These technologies have been widely deployed on real-world platforms. This report will share the key challenges we encountered during our research and the solutions we provided.