科学研究

Jianping Li received the B.S. degree in geographic information system (GIS) and the Ph.D. degree in photogrammetry and remote sensing from Wuhan University, Wuhan, China, in 2015 and 2021, respectively. He is currently a Research Fellow with the School of Electrical and Electronic Engineering, Nanyang Technological University, Singapore. His research interests include 3-D sensing system integration, uncrewed aerial vehicle (UAV)/uncrewed ground vehicle (UGV) mapping, robot perception, and point cloud data processing, with over 50 publications in venues, such as ISRPS, IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS, IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, RAL, ICRA, and IROS. He received the Science and Technology Progress Award in Surveying and Mapping in 2019 and 2023, the National Excellent Doctoral Thesis on LiDAR in 2022, and the First Prize of the 20-th China International Industrial Expo in 2020.

内容摘要：

Accurate and adaptive 3D perception in complex and unexposed environments is essential for digitalization, infrastructure inspection, and autonomous mapping. Yet traditional LiDAR-based systems rely on fixed scanning patterns and passive data acquisition, limiting their efficiency and robustness in dynamic or cluttered scenes. This talk presents a comprehensive research framework that progresses from active sensor control and hardware design to large-scale mapping and reconstruction, ultimately enabling intelligent perception for complex spatial environments.The first part introduces an active LiDAR platform capable of dynamically adjusting its scanning behavior in response to scene geometry. Both the mechanical design and control algorithms are developed to actively enhance environmental observability and maintain stable localization. Building on this foundation, predictive and uncertainty-aware control strategies are formulated by integrating model-based optimization with data-driven decision making, allowing the sensor to anticipate and adapt to future perception challenges.The second part focuses on building large-scale and consistent 3D maps from the actively acquired data. A graph-based optimization framework is developed to maintain global consistency across multiple sessions while preserving local geometric accuracy. For lightweight and wearable systems, continuous-time optimization techniques are introduced to handle asynchronous LiDAR and inertial measurements. In parallel, reinforcement-driven adaptive meshing methods are designed to achieve real-time 3D reconstruction and monitoring in unexposed or occluded environments.

The talk concludes with a discussion on the application of these methods in complex spatial settings, such as underground infrastructure, construction sites, and collaborative multi-robot mapping.

联系人：智能系统与人形机器人国际研究中心，潘婷婷

编辑：袁晓慧