The low-altitude economy heavily relies on new carriers represented by unmanned aerial vehicles (UAVs). The localization accuracy of UAVs highly relies on the Global Navigation Satellite System (GNSS), which can be easily affected in low-altitude urban environments, making it difficult to maintain effective localization accuracy. To solve this problem, this paper proposes a UAV autonomous localization method with keypoints on the edges of semantic objects (KESO). Firstly, semantic objects within the working area are selected, and then the latitude, longitude, and altitude of these semantic objects’ keypoints are measured to construct a database. By identifying the semantic objects from aerial images and detecting the edge of the semantic objects, the keypoints of the semantic objects are obtained. Finally, by matching the detected keypoints in the aerial images with the keypoints in the database, the UAV’s position can achieve a high-precision position when satellite signals are blocked in low-altitude urban environments. As verified by real flight data, the results show that the localization error is less than 5 m, and the edges of objects can obtain more accurate keypoints to help UAVs locate more accurately. This paper can provide a reference for UAV localization in the urban environments of the low-altitude economy.

Because of the resilient frame structure, the factor graph is often used in navigation systems to solve the sensor asynchrony problem and realize plug-and-play effectively in the navigation information fusion method. To improve the fault detection performance of resilient integrated navigation systems under complex interference environments, a fault detection method in factor graph navigation framework based on INS measurements and GNSS pseudo-range measurements is proposed in this paper. The proposed method can effectively locate the fault satellite pseudo-range information based on the Chi-square fault detection method. Due to the plug-and-play characteristic of the factor graph framework, this method can quickly isolate faults to improve navigation accuracy. Finally, the effect of the method is verified by simulation and experiment. Compared with the Chi-square fault detection method, positioning accuracy is improved by more than 40%.

Accurate detection of road surface information is crucial for enhancing vehicle driving safety and ride comfort. To overcome the limitation that traditional suspension systems struggle to respond to road excitations in real time due to time delays in signal acquisition and control, suspension preview control technology has attracted significant attention for its proactive adjustment capability, with efficient road surface information perception being a critical prerequisite for its implementation. This paper systematically reviews road surface information detection technologies for suspension preview, focusing on the identification of potholes and speed bumps. Firstly, it summarizes relevant publicly available datasets. Secondly, it sorts out mainstream detection methods, including traditional dynamic methods, 2D image processing, 3D point cloud analysis, machine/deep learning methods, and multi-sensor fusion methods, while comparing their applicable scenarios and evaluation metrics. Furthermore, it emphasizes the core role of elevation information (e.g., pothole depth, speed bump height) in suspension preview control and summarizes elevation reconstruction technologies based on LiDAR, stereo vision, and multi-modal fusion. Finally, it prospects future research directions such as optimizing robustness, improving real-time performance, and reducing labeling costs. This review provides technical references for enhancing the accuracy of road surface information detection and the control efficiency of suspension preview systems, and it is of great significance for promoting the development of intelligent chassis.

This paper investigates the potential of utilizing frequency-varying negative stiffness properties of the inerter in vehicle suspensions to improve the comprehensive dynamic performance. Firstly, the dynamic characteristics of the air spring and the inerter are analyzed to evaluate the feasibility of integrating them into vehicle suspensions. Secondly, a quarter-vehicle model incorporating the nonlinear stiffness features of an air spring is established, and a vehicle semi-active control strategy for the air inerter–spring–damper (ISD) suspension based on the frequency-varying negative stiffness of the inerter is proposed. Thirdly, in order to implement this strategy effectively, this paper builds an uneven road surface estimator based on the discrete Kalman filter with unknown input (DKF-UI) and an uneven road surface frequency identifier based on the first order-zero crossing algorithm. Finally, the superiority of the proposed suspension system is verified by simulations, and the results reveal that with respect to the passive air suspension, the peak values of the gains of the body acceleration, the suspension working space and the dynamic tire load are reduced. The vehicle semi-active air ISD suspension exhibits reductions of 68.7%, 51.0%, 67.6% in the body frequency region and 5.1%, 9.9%, 7.9% in the wheel frequency region, respectively. Under a segment sinusoidal road input, the RMS values of the three performance indicators exhibit reductions of 79.2%, 57.8%, 77.7% at 1.4 Hz and 6.8%, 15.4%, 5.8% at 11 Hz, respectively. Consequently, the vehicle semi-active air ISD suspension proposed in this paper has significant performance improvement at both the vehicle body natural frequency and the wheel natural frequency region compared to the passive air suspension, which indicates that it has better ride comfort and road holding performance.