In the field of autonomous land vehicles (ALVs), environmental perception serves as the gateway for ALVs to acquire external information, directly determining the capabilities and intelligence level of autonomous driving. In the research of environmental perception technology for ALVs in field environments, achieving real-time, accurate, and robust obstacle detection is fundamental to realizing autonomous navigation. However, existing research often ignores the detection of negative obstacles, making the detection of positive and negative obstacles crucial for the safe navigation of ALVs. Therefore, we propose a method for ALV to use dual-LiDAR for positive and negative obstacles detection. The ground segmentation suffers from issues such as scene-specificity and threshold dependence. This method employs a horizontally mounted LiDAR to achieve ground segmentation and positive obstacle detection utilizing an adaptive threshold ground point cloud segmentation method with an improved ray method. Subsequently, a vertically mounted LiDAR is used to detect negative obstacles through local analysis of point clouds and clustering of feature point pairs to achieve local obstacle avoidance. A series of experiments were conducted in multiple field environments, and the results demonstrate that the proposed method achieves a successful detection rate of 98.5% for positive obstacles and 98.7% for negative obstacles in different field environments.

Semantic segmentation of aerial images is crucial in various practical applications, encompassing traffic management, search tasks, urban planning, and more. However, due to the unique shooting angles of aerial images, there are significant challenges in accurately segmenting objects, including large variations in object scales, deformations, and unclear features of small targets. To address this, we propose a multi-level feature mining network based on an encoder-decoder architecture called MLFMNet, aimed at excavating and integrating multi-level feature information in aerial images to enhance segmentation accuracy and robustness. MLFMNet leverages skip connections to obtain hierarchical feature representations from the encoder. Subsequently, through learnable fusion module (LFM) and feature reconstruction module (FRM) in the proposed decoder, it progressively fuses and reconstructs these features, thereby achieving accurate semantic segmentation. To tackle issues of significant size variations and deformations in objects, we design an irregular pyramid receptive field (IPRF) module embedded at the bottom of the encoder to capture receptive fields from multiple feature vectors, thus further mining abstract features. Moreover, to address the challenge of low segmentation and detection accuracy for small targets, a fine-grained feature mining (FGFM) module is embedded in the bottom of the decoder to capture spatial detail features. Particularly, MLFMNet-B achieves an mIoU of 70.8%, ranking fourth in the official leaderboard of the UAVid test set. The code is available at https://github.com/wxy16/MLFMNet .

Accurate mathematical modeling of state of charge (SOC) prediction is essential for battery management systems (BMSs) to improve battery utilization efficiency and ensure a good safety performance. The current SOC prediction framework only considers battery-related features but ignores vehicle information. Additionally, in light of the emergence of time-series Transformers (TSTs) that harness the power of multi-head attention, developing a SOC prediction model remains a significant challenge. Therefore, we introduce a new framework that integrates laboratory battery data with mathematical vehicle model features to improve the accuracy of the SOC and propose a prediction model named DLPformer, which can effectively capture variations in the SOC attributed to both trend and seasonal patterns. First, we apply Matlab/Simulink to simulate a mathematical model of electric vehicles and process the generated vehicle data with Spearman correlation analysis to identify the most relevant features, such as the mechanical losses of the electric motor, differential, and aerodynamic drag. Then, we employ a data fusion method to synchronize the heterogeneous datasets with different frequencies to capture the sudden changes in electric vehicles. Subsequently, the fused features are input into our prediction model, DLPformer, which incorporates a linear model for trend prediction and patch-input attention for seasonal component prediction. Finally, in order to effectively evaluate the extrapolation and adaptability of our model, we utilize different driving cycles and heterogeneous battery datasets for training and testing. The experimental results show that our prediction model significantly improves the accuracy and robustness of SOC prediction under the proposed framework, achieving MAE values of 0.18% and 0.10% across distinct driving cycles and battery types.

Navigation in unknown 3D environments aims to progressively find an efficient path to a given target goal in unseen scenarios. A challenge is how to explore the navigation quickly and effectively. An end-to-end learning approach has been proposed to extract geometric shapes from RGB images, but it is not suitable for large environments due to its exhaustive exploration with exponential search space. Active Neural SLAM (ANS) presents a Neural SLAM module to maximize the exploration coverage to tackle the active SLAM task. However, ANS still frequently visits the explored areas due to the inappropriate local target selection. In this paper, we propose a Neural Attention-based Visual SLAM (NAVS) model to explore unknown 3D environments. Spatial attention is provided to quickly identify obstacles (such as similarly colored tea table or floor). We also leverage the priority of unknown regions in the short-term goal decision to avoid frequent exploration with a channel attention. The experimental results show that our model can build a more accurate map than ANS and other baseline methods with less running time. In terms of relative coverage, NAVS achieves a 0.5 $$\%$$% improvement over ANS in overall and a 1.1 $$\%$$% improvement over ANS in large environments.

Precise localisation and navigation are the two most important tasks for mobile robots. Visual simultaneous localisation and mapping (VSLAM) is useful in localisation systems of mobile robots. The wide‐angle camera has a broad field of vision and more abundant information on images, so it is widely used in mobile robots, including legged robots. However, wide‐angle cameras are more complicated than ordinary cameras in the design of visual localisation systems, and higher requirements and challenges are put forward for VSLAM technologies based on wide‐angle cameras. In order to resolve the problem of distortion in wide‐angle images and improve the accuracy of localisation, a sampling VSLAM based on a wide‐angle camera model for legged mobile robots is proposed. For the predictability of the periodic motion of a legged robot, in the method, the images are sampled periodically, image blocks with clear texture are selected and the image details are enhanced to extract the feature points on the image. Then, the feature points of the blocks are extracted and by using the feature points of the blocks in the images, the feature points on the images are extracted. Finally, the points on the incident light through the normalised plane are selected as the template points; the relationship between the template points and the images is established through the wide‐angle camera model, and the pixel coordinates of the template points in the images and the descriptors are calculated. Moreover, many experiments are conducted on the TUM datasets with a quadruped robot . The experimental results show that the trajectory error and translation error measured by the proposed method are reduced compared with the VINS‐MONO, ORB‐SLAM3 and Periodic SLAM systems.