Trustworthy Perception and Control for Unmanned Autonomous Systems (UAS): From Unmanned Ground Vehicles to Unmanned Aerial Vehicles

Abstract

Unmanned autonomous systems (UAS), such as unmanned ground vehicles (UGV) and unmanned aerial vehicles (UAV), are one of the key “ingredients” for the arrival of smart cities. We have witnessed great advancement of the UAS in the past decades. For example, the demonstration of the UGV in constrained areas was available, which has the potential to alleviate the problems of traffic congestion and accidents. The UAV for autonomous parcel delivery can further unlock the 3D space for the betterment of human beings. However, we rarely see the massive deployment of UAS in complex urban environments, because of the limitation of existing UAS technology. For example, perception (positioning and mapping, etc) and control are significantly challenged in urban scenarios, leading to unacceptable uncertainty in UAS. Trustworthiness, which is of great importance to the UAS, is hard to guarantee due to the complex environmental conditions. In fact, the development of UAS technology is getting into the “long-tail period” which is an internationally recognized challenge towards the massive deployment of the UAS in our daily life. This talk will present some recent research findings on the advocacy of the trustworthiness of the UAS perception and control, which include: (1) the outlieraware LiDAR-centered multi-sensory integration-based high-accuracy mapping for UGV in complex urban areas; (2) visual localization within the prior light-weight map for UAV; (3) the positioning-control joint optimization model for the resilient UAV, which exploits the complementariness of the flight dynamics and the positioning state estimation; (4) safety certification for the visual/LiDAR localization of UAV/UGV; (5) vision aided RTK positioning for UAV parcel delivery in complex urban canyons; (6) UAV-aided additive manufacturing without spatial limitation; and (7) intelligent roadside infrastructure and UGV collaboration for high definition map update for UGV. Many of the codebases have been open-sourced for the benefit of the research community or knowledge transferred for industry collaboration for far-reaching and impactful outcomes.

Speaker

Dr Weisong Wen received a PhD degree in Mechanical Engineering from The Hong Kong Polytechnic University (PolyU) in 2020 and was also a visiting PhD student at the Faculty of Engineering, University of California, Berkeley (UC Berkeley) in 2018. Having served as a Postdoc Researcher at the PolyU for a short period, Dr Wen is now serving as a Research Assistant Professor at the Department of Aeronautical and Aviation Engineering of PolyU. His research interests focus on trustworthy spatial perception (positioning, mapping, etc) and control (ground vehicle control and aerial vehicle control, etc) for unmanned autonomous systems (UAS), such as the unnamed ground vehicles (UGV) and unmanned aerial vehicles (UAV). Dr Wen has published more than 30 SCI journal papers and 32 international conference papers (Google Citations: 907, h-index:18 by 1st June 2023), within the field of UAS and intelligent transportation systems. He has served as the branch/field chairman of IEEE ITSC 2022/2023, the top conference in the field of intelligent transportation, ICGNC 2022, and ION GNSS+ 2022/2023, the annual meeting of the American Navigation Association. He regularly organized workshops in IEEE ITSC 2022/2023 on the topic of trustworthy UAS (UAV, UGV, etc) in complex urban canyons by collaborating with researchers from MIT, RWTH Aachen University, HKUST, University of Macau, Technische Universität Chemnitz, German Aerospace Center (DLR), and Tsinghua University, etc. Meanwhile, He served as the Young Editor Board Member and Leading Guest Editor of several journals. He lead the open-sourced UrbanLoco dataset together with the Mechanical Systems Control (MSC) lab (led by Prof. Masayoshi Tomizuka) at UC Berkeley in 2020. He also open-sourced the UrbanNav dataset and the GraphGNSSLib library for reliable positioning in urban canyons. His research on high-precision LiDAR-based mapping algorithms for UGV in urban environments won the ION GNSS+ 2020 Best Presentation Award, TechConnect World Innovation Conference, and Expo Innovation Award 2021, and was recently patented in the U.S. He also won the First Prize in Hong Kong Section in Qianhai-Guangdong-Macao Youth Innovation and Entrepreneurship Competition, 2019. His research in vision-aided UAV RTK positioning received a research award from Meituan UAV with the potential for autonomous parcel delivery. In addition, his recent research on UAV-aided additive manufacturing recently received funding support from the PolyU Research Institute for Advanced Manufacturing (RIAM). His research work was recognized internationally which led to the regular explicit PhD student exchange from international institutions, such as The Technische Universität Chemnitz in Germany, Leibniz University Hannover in Germany, and Tokyo University of Marine Science and Technology in Japan to Dr Wen’s research group. Dr Wen’s research received multiple research funding from both academics and industry fields, such as Hong Kong Innovation and Technology Commission (ITF), PolyU, Huawei Technologies, Meituan UAV, and Tencent, leading to more than 7.5 million HKD in research funds as PI since 2021.