AASS Seminar: Jianhao Jiao, UCL

Who: Dr. Jianhao Jiao, a Senior Research Fellow at University College London (UCL)
What: External seminar with title "SLAM for Robotics: Lifelong Spatial Memory and Navigation in Unstructured Worlds"
When: December 4, 13:00-14:00
Where: Hörsal T, Teknikhuset

Abstract: Modern robotics demands that autonomous systems, from agile quadrupeds to humanoids, operate reliably in dynamic field environments to assist people in complex tasks like logistics and inspection. This utility fundamentally relies on the robot possessing a reliable, lifelong spatial memory. Traditional Simultaneous Localization and Mapping (SLAM) systems use dense, high-fidelity geometric maps that are non-scalable, storage-intensive, and inherently fragile when confronted with real-world dynamism and temporal change. 

To overcome this crisis of scalability, this talk presents a transition towards a sparse, cognitive spatial memory architecture. We introduce OpenNavMap, a structure-free topometric mapping approach that abandons explicit 3D geometry in favor of a lightweight, multi-layered graph representation. This system achieves metric-level consistency, essential for planning, by dynamically reconstructing local 3D structure on-demand via neural inference, enabling robust cross-session map merging. This novel paradigm has been validated across diverse real-world scenarios, including deployment on mobile robotic platforms. The discussion will also conclude by focusing on the critical future challenge: enabling this spatial memory mechanism to achieve lifelong adaptability and real-time maintenance in highly unstructured and visually challenging domains, specifically addressing navigation in  forests.

Bio: Dr. Jianhao Jiao is a Senior Research Fellow at University College London (UCL), holding a Ph.D. in Electronic and Computer Engineering from the Hong Kong University of Science and Technology. His research specializes in SLAM, sensor fusion, and robust robotic navigation, exemplified by pioneering works such as M-LOAM, FusionPortable dataset, and the scalable, structure-free visual navigation system, OpenNavMap. Dr. Jiao’s ultimate research objective is to endow autonomous systems with lifelong, cognitive spatial memory mechanisms capable of dynamic updating, directed towards applications in challenging, unstructured environments such as subterranean mines and  forests.