Feng Huang

GNSS real-time kinematic (RTK) and LiDAR odometry are complementary, offering global and local positioning capabilities, respectively. GNSS-RTK/LiDAR integration primarily focuses on two aspects: (1) 3D LiDAR-aided (3DLA) GNSS non-line-of-sight (NLOS) mitigation using point cloud data. (2) State estimation via tightly-coupled GNSS/LiDAR integration, which improves GNSS geometry, yielding better float solutions and lower estimation uncertainty. Combining 3DLA NLOS mitigation with tightly-coupled integration can thus enhance urban positioning reliability. However, the current 3DLA GNSS-RTK system remains limited in terms of NLOS detection reliability, tightly-coupled integration effectiveness, and computational efficiency. This paper proposes a tightly-coupled GNSS-RTK/LiDAR/INS integration system with 3DLA NLOS mitigation. We focus on improving the robustness of 3DLA NLOS detection, the efficiency and effectiveness of the tightly-coupled GNSS-RTK/LiDAR integration. Evaluations on the challenging UrbanNav dataset show over 70% improvement in positioning accuracy compared to conventional GNSS-RTK, achieving submeter to meter-level precision. Runtime analysis confirms the system supports real-time optimization and ambiguity resolution with an average computation time of 70 ms.

GNSS/INS/LiDAR-based Multi-Sensor Fusion (MSF) systems facilitate the efficient integration of multiple navigation sensors to deliver stable and dependable positioning outcomes in Autonomous Vehicles (AVs). Generally, AVs are bound to operate in diverse scenarios. However, state-of-the-art spoofing attack algorithms can identify vulnerable periods of MSF systems and execute aggressive GNSS spoofing in certain specific scenarios with high success rates. The present study proposes a security defense algorithm for MSF systems based on an enhanced Kalman filter, incorporating the Measurement Variance Monitoring (MVM) and Minimum Constraint of Covariance Matrix (MCCM) model. MVM effectively constrains the impact of GNSS signals with large positioning errors, thereby mitigating the adverse effects caused by spoofing attacks. MCCM fully leverages LiDAR information between two consecutive GNSS signals to enhance the credibility of LiDAR measurements, thus improving the positioning correction capability. Furthermore, considering the diverse application environments of AVs, this paper primarily focuses on sensor uncertainty variation and validates different models in varying scenarios. Ultimately, real-world data demonstrate the effectiveness of the proposed anti-spoofing model, which can significantly mitigate positioning errors caused by GNSS spoofing attacks without affecting navigation accuracy.

Leveraging multiple sensors enhances complex environmental perception and increases resilience to varying luminance conditions and high-speed motion patterns, achieving precise localization and mapping. This paper proposes, ECMD, an event-centric multisensory dataset containing 81 sequences and covering over 200 km of various challenging driving scenarios including high-speed motion, repetitive scenarios, dynamic objects, etc. ECMD provides data from two sets of stereo event cameras with different resolutions (640*480, 346*260), stereo industrial cameras, an infrared camera, a top-installed mechanical LiDAR with two slanted LiDARs, two consumer-level GNSS receivers, and an onboard IMU. Meanwhile, the ground-truth of the vehicle was obtained using a centimeter-level high-accuracy GNSS-RTK/INS navigation system. All sensors are well-calibrated and temporally synchronized at the hardware level, with recording data simultaneously. We additionally evaluate several state-of-the-art SLAM algorithms for benchmarking visual and LiDAR SLAM and identifying their limitations. The dataset is available at https://arclab-hku.github.io/ecmd/.

Light detection and ranging (LiDAR) can provide continuous and stable pose estimation with the model of Normal Distribution Transform (NDT), which is widely used in Autonomous Vehicles (AVs), even under adverse weather conditions. However, there are few studies about the influence of inclement weather on LiDAR positioning results. In this paper, different weather scenarios (rain, fog, and snow) are composed with synthetic LiDAR datasets based on state-of-the-art weather simulators. Then, the impacts of different adverse weather conditions are quantitatively evaluated in terms of positioning accuracy and uncertainty. Afterward, we perform the first study to qualitatively analyze the relationship between meteorological weather standards and LiDAR positioning performances, which is significant but unexplored. Evaluated results indicate that NDT matching performance will deteriorate in adverse weather conditions, especially when the meteorological level is ‘Heavy’ or ‘Violent’, threatening the AVs’ positioning security seriously. Therefore, Results of this paper provide more basis for the realization of high-precision positioning in adverse weather conditions, to ensure the positioning safety of AVs.

The inherent vulnerability of the Global Navigation Satellite System (GNSS) leads to the ease of implementation of spoofing attacks. The latest GNSS spoofing attack schemes still suffer from low success rate, long attack time, and poor concealment. To improve the success rate, an efficient GNSS spoofing attack method for a vehicle-mounted Multi Sensors Fusion (MSF) system is proposed based on the scenario classification models with a spatial database. Firstly, the influence of the two typical urban scenarios, which are 1) the road with buildings on both sides and 2) tunnels, on the GNSS spoofing attack is analyzed. Then a dynamic Bayesian network model considering the sky visibility generated with the 3D building models and tunnel models inside the spatial database is established to quantify the difficulty of the attack. Furthermore, the scenarios of the victim can be classified into high-risk and low-risk scenarios. When the vehicle is just out of the tunnel or in open scenarios, attackers can select these high-risk scenarios and implement aggressive spoofing attacks. Then the efficiency of the GNSS spoofing attack can be significantly improved. Finally, the proposed attack scheme is demonstrated by actual world data with simulated spoofing attacks in urban areas.

Hostile spoofing attacks on the Global Navigation Satellite System (GNSS) receiver increase the risk of catastrophic consequences to autonomous driving systems. This paper addresses the problem of vulnerability of Kalman filter (KF) under spoofing attack. A state-of-the-art spoofing attack method based on maximizing the lateral deviation is utilized for verification and analysis. To analyze the vulnerability in actual road scenarios better, an analytic error model of the mechanism of GNSS spoofing is derived. Except for the uncertainty of the initial MSF state, the uncertainty of the uncertainty of Light Detection And Ranging (LiDAR) and the uncertainty of GNSS, a new factor in spoofing attack, the update frequency of different sensors, is investigated in the paper, which in fact is a key factor to increase the immunity Multi-Sensors Fusion (MSF) systems. Experiments were performed in a typical urban scenario of the KAIST dataset. When the update frequency ratio between GNSS and LiDAR is 1, 2, 5, and 10, successful spoof attacks can be performed if the Standard Deviation (STD) of GNSS is smaller than 4m, 2.7m, 1.1m, and 0.7m respectively. Therefore, experiments confirm that the uncertainty of GNSS and the update frequency ratio between LiDAR and GNSS are critical for spoofing attacks, which provides an indication for designing a defense strategy in the future.

Global Navigation Satellite System Real-time Kinematic (GNSS-RTK) is an indispensable source for the absolute positioning of autonomous systems. Unfortunately, the performance of the GNSS-RTK is significantly degraded in urban canyons, due to the notorious multipath and Non-Line-of-Sight (NLOS). On the contrary, LiDAR/inertial odometry (LIO) can provide locally accurate pose estimation in structured urban scenarios but is subjected to drift over time. Considering their complementarities, GNSS-RTK, adaptively integrated with LIO was proposed in this paper, aiming to realize continuous and accurate global positioning for autonomous systems in urban scenarios. As one of the main contributions, this paper proposes to identify the quality of the GNSS-RTK solution based on the point cloud map incrementally generated by LIO. A smaller mean elevation angle mask of the surrounding point cloud indicates a relatively open area thus the correspondent GNSS-RTK would be reliable. Global factor graph optimization is performed to fuse reliable GNSS-RTK and LIO. Evaluations are performed on datasets collected in typical urban canyons of Hong Kong. With the help of the proposed GNSS-RTK selection strategy, the performance of the GNSS-RTK/LIO integration was significantly improved with the absolute translation error reduced by more than 50%, compared with the conventional integration method where all the GNSS-RTK solutions are used.