Simultaneous Localization and Mapping (SLAM) is one of the fundamental capabilities for autonomous vehicles to achieve accurate localization in dynamic urban environments. However, in real-world scenarios, the presence of moving objects such as pedestrians, bicycles, or vehicles affects the localization accuracy of SLAM and leaves behind ghost trails in the created map. Hence, it is essential for SLAM to remove dynamic objects in real-time to improve its accuracy. In this paper, we present DOR-LINS, a real-time moving objects removal framework for in LiDAR-Inertial SLAM. Built on the foundation of LIO-SAM, DOR-LINS integrates the capability of removing dynamic objects, enabling it to remove most of the dynamic objects and achieve precise odometry. Our contributions are reflected in three aspects. Firstly, our method extracts ground points from the current scan and clusters non-ground points to segment them into static and potential dynamic clusters. Then, we divide the submap and current scan into multiple bins and introduce a novel concept called ground pseudo occupancy to describe the occupancy of each bin. Secondly, based on the ground pseudo occupancy , dynamic clusters, and static clusters, we propose an approach named Beam Tracing Test (BTT) and combine it with Scan Ratio Test (SRT) to select candidate dynamic bins. Finally, we employ a dynamic point verification algorithm to filter out actual dynamic points from these candidate dynamic bins. As experimentally evaluated on the UrbanLoco dataset, our proposed method removes many dynamic points and yields promising performance against state-of-the-art methods.

This paper summarized the research status, imaging model, systems calibration, distortion correction, and panoramic expansion of panoramic vision systems, pointed out the existing problems and put forward the prospect of future research. According to the research status of panoramic vision systems, a panoramic vision system with single viewpoint of refraction and reflection is designed. The systems had the characteristics of fast acquisition, low manufacturing cost, fixed single-view imaging, integrated imaging, and automatic switching depth of field. Based on these systems, an improved nonlinear optimization polynomial fitting method is proposed to calibrate the monocular HOVS, and the binocular HOVS is calibrated with the Aruco label. This method not only improves the robustness of the calibration results, but also simplifies the calibration process. Finally, a real-time method of panoramic map of multi-function vehicle based on vcam is proposed.

LiDAR-binocular camera systems have attracted increasing research attention with their advantages of two sensors. The two-sensor data fusion premise accurately calibrates the external parameters of both sensors. However, existing LiDAR and binocular camera calibration uses geometric methods to calibrate the left and right cameras with the LiDAR or register the LiDAR point cloud and camera depth map to obtain the extrinsic parameters. Existing methods have difficulty obtaining low binocular camera baseline length and reprojection errors simultaneously. This paper proposes a binocular camera and LiDAR extrinsic parameter calibration method based on a BP neural network and checkerboard. First, a LiDAR-binocular camera calibration model is established, and the extrinsic parameters and intermediate variables to be calibrated are determined. The calibration rotation matrix has multiple variables that are not independent of each other, creating a complex model. The nine rotation matrix variables are decoupled and reduced using Euler angle representation. Then, the BP neural network structure is determined using the external parameter calibration data characteristics. The calibration algorithm accuracy is improved by introducing the baseline and reprojection errors into the loss function. Finally, we use real data to check the calibration performance and introduce the baseline length and reprojection errors as indicators to improve the calibration result evaluation accuracy and effectiveness. The results show that the proposed calibration method has good consistency and small baseline and reprojection errors. Compared with traditional PnP methods, the average baseline error is reduced by 64.804%, and the reprojection error is reduced by approximately 10%.

The extrinsic calibration of visual-inertial sensors has always been a complex problem for autonomous vehicles because the vehicle usually moves on a plane, rendering the calibration problem degenerate. Even if the vehicle drives on a slope, many great works are ineffective due to insufficient excitation. To address these challenges, we propose a method for calibrating the extrinsic parameters of camera and IMU sensors for vehicles, which involves two key contributions. The first is the introduction of a virtual vehicle reference frame, acting as a bridge to calibrate rotation in degenerate case. This involves formulating two least-squares problems to estimate the 3D relative rotation of the camera-IMU. Secondly, we propose an approach for estimating camera-IMU translation when a vehicle is navigating on a slope. This is done by solving a linear system, and these values are then refined using an algorithm that takes into account the gravitational magnitude. Extensive experimental results show that our proposed method effectively estimates rotation in planar motion and translation in sloping motion. Moreover, it achieves competitive accuracy and consistency when compared to state-of-the-art approaches.

Airborne capture net devices mounted on unmanned aerial vehicles (UAVs) provide an effective physical countermeasure against rogue drones. However, conventional deterministic designs neglect parameter uncertainties such as manufacturing tolerances and assembly errors that degrade flexible net deployment dynamics. To resolve this, this study proposes a multi-objective robust design optimization framework integrating a feedforward neural network (FNN) with uncertainty quantification. A lumped-mass dynamic model is formulated; static deployment experiments validate its high fidelity during the initial phase (t ≤ 0.08 s), while subsequent full-deployment and contraction phases are captured via experimentally calibrated numerical extrapolation. To mitigate computational costs, an adaptive FNN surrogate model is developed, reducing single-evaluation times from hours to seconds. Monte Carlo simulations coupled with the non-dominated sorting genetic algorithm II (NSGA-II) are then employed to optimize the mean and standard deviation of deployment and effective distances. Compared with deterministic optimization, the robust solution yields a 23.8% increase in the 95th percentile of deployment distance with only a 1.5% decrease in effective distance. Crucially, the standard deviations of these metrics decrease by 22.5% and 38.5%, respectively, demonstrating significantly attenuated sensitivity to parameter variations and offering a valuable engineering framework for robust airborne interception.