Hongji Liu

I am a Ph.D. candidate from The Hong Kong Univerisity of Science and Technology at the Robot Motion Planning and Control Lab under the supervison of Prof. Jun Ma, and also at the Robotics and Multi-perception Lab in Robotics Institute, under the supervision of Prof. Ming Liu and Prof. Qifeng Chen. I received my bachelor's degree in Software Engineering from Sun Yat-sen University in 2019.

My research interests mainly include high-definition maps, semantic mapping, robot state estimation, and SLAM (Simultaneous Localization and Mapping), mainly applied to autonomous driving and robotics.

Maps play a key role in indoor robots, autonomous vehicles and other robotics derived applications. Maps with complete functions, easy storage and high accuracy are very important for robot perception, localization and planning. Therefore, I am devoted to exploring efficient and accurate map expression forms applicable to robots in different application scenarios and tasks. It is expected to take the map as the link to help improve the performance of the robot in different application scenarios.

Selected publications

Enhance Inverse Perspective Mapping for Automatic Vectorized Road Map Generation

Hongji Liu, Linwei Zheng, Yongjian Li, Mingkai Tang, Xiaoyang Yan, and Jun Ma

Under Review , 2025

As unmanned ground vehicles (UGVs) applications expand to large-scale and open road scenarios, vectorized high-definition maps (VHD maps) demonstrate greater potential in solving state correction problems than traditional metric maps. Previous related studies typically employ proprietary versions of VHD maps without fully leveraging the common traffic elements' information. In addition, there is a lack of research on efficient interaction methods between UGVs and VHD maps. To fill these gaps, we propose a universal UGV state estimation system and a query-based VHD map data exchange protocol. The system utilizes VHD maps to correct the lateral and longitudinal positions as well as the yaw orientation of UGVs. The data exchange protocol enables UGVs to obtain real-time VHD map information and process it efficiently. To ensure universality, we accommodate two widely used VHD map formats, ASAM OpenDRIVE and Apollo OpenDRIVE and provide corresponding map parsing methods. The evaluation of the system is conducted both in simulated and real-world scenes. In the simulation experiments, we fully measure the effectiveness and accuracy of our method, as well as its sensitivity to measurement noise. In real-world experiments, we compare the state estimation accuracy of our system with SOTA simultaneous localization and mapping methods on an open road. The results show that our system demonstrates better accuracy than other baselines on most data sequences. The proposed map data exchange protocol meets real-time requirements. For the community's reference, the system code is available at https://github.com/liuhj86/VHDMap-SE.
VHDMap-SE: A Universal Vectorized High-definition Map Aided Vehicle State Estimator

Hongji Liu, Mingkai Tang, Mingkai Jia, Yingbing Chen, Jin Wu and Ming Liu

IEEE Transactions on Intelligent Vehicles (T-IV) , 2024

Abs PDF Website

As unmanned ground vehicles (UGVs) applications expand to large-scale and open road scenarios, vectorized high-definition maps (VHD maps) demonstrate greater potential in solving state correction problems than traditional metric maps. Previous related studies typically employ proprietary versions of VHD maps without fully leveraging the common traffic elements' information. In addition, there is a lack of research on efficient interaction methods between UGVs and VHD maps. To fill these gaps, we propose a universal UGV state estimation system and a query-based VHD map data exchange protocol. The system utilizes VHD maps to correct the lateral and longitudinal positions as well as the yaw orientation of UGVs. The data exchange protocol enables UGVs to obtain real-time VHD map information and process it efficiently. To ensure universality, we accommodate two widely used VHD map formats, ASAM OpenDRIVE and Apollo OpenDRIVE and provide corresponding map parsing methods. The evaluation of the system is conducted both in simulated and real-world scenes. In the simulation experiments, we fully measure the effectiveness and accuracy of our method, as well as its sensitivity to measurement noise. In real-world experiments, we compare the state estimation accuracy of our system with SOTA simultaneous localization and mapping methods on an open road. The results show that our system demonstrates better accuracy than other baselines on most data sequences. The proposed map data exchange protocol meets real-time requirements. For the community's reference, the system code is available at https://github.com/liuhj86/VHDMap-SE.
PGO-IPM: Enhance IPM Accuracy with Pose-guided Optimization for Low-cost High-definition Angular Marking Map Generation

Hongji Liu, Linwei Zheng, Xiaoyang Yan, Zhenhua Xu, Bohuan Xue, Yang Yu, and Ming Liu

35th IEEE Intelligent Vehicles Symposium (IV) , 2024

Abs PDF Website

High-definition angular marking maps (HDAM maps) are vital in large-scale environments with variable appearances. In these scenarios, unmanned ground vehicles (UGVs) can use angular markings for localization because they are easy to identify and informative for localization. However, creating such a marking map relies heavily on manual measurement and annotation, which is time-consuming and laborious. Although Inverse Perspective Mapping (IPM) offers a low-cost and automated alternative, its accuracy is compromised by vehicle motion and the arduous pre-calibration of the IPM matrix. To fill these gaps, we propose a pose-guided optimization framework for IPM. This framework enables the automated generation of HDAM maps, while concurrently refining the preliminary IPM matrix. We deployed the proposed method in two different automated ports, and the method yielded HDAM maps with near-centimeter precision. Moreover, the refined IPM matrix matched the accuracy of manual calibrations. The supplementary materials and videos are available at http://liuhongji.site/PGO-IPM/.
360ST-Mapping: An Online Semantics-Guided Topological Mapping Module for Omnidirectional Visual SLAM

Hongji Liu, Huajian Huang, Sai-Kit Yeung, and Ming Liu

IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) , 2022

Abs PDF Website

Topological map as an abstract representation of the observed environment has the advantage in path planning and navigation. Here we proposed an online topological mapping method, 360ST-Mapping, by making use of omnidirectional vision. The 360° field of view allows the agent to obtain consistent observation and incrementally extract topological environment information. Moreover, we leverage semantic information to guide topological places recognition further improving performance. The topological map possessing semantic information has the potential to support semantics-related advanced tasks. After combining the topological mapping module with the omnidirectional visual SLAM, we conduct extensive experiments in several large-scale indoor scenes to validate the effectiveness.

SEE ALL PUBLICATIONS