Search

Light detection and ranging (LiDAR)-based simultaneous localization and mapping (SLAM), known for its precision and resilience to interference, has been widely utilized in open and unknown environments. However, the common LiDAR sensors (e.g., VLP-16) face many limitations such as a small vertical field of view, hindering the provision of adequate vertical motion constraints, which leads to easier error accumulations in the vertical direction and a significant hurdle for deployment in large-scale scenarios. Motivated by this problem, a covisibility region aware 3D-LiDAR SLAM method, is proposed based on generative road surface model and long-term association. First, different from existing methods that often overlook the ground points, the role of ground points is reexamined from a fresh perspective, and the intrinsic relationship between the ground points and the sequential states is established, which specifically compensates for the inadequacy of vertical pose constraints. Then, a pipeline for processing ground points within generalized terrains is proposed, and a long-term pose constraint (denoted as ROSE-constraint) for region-level tracking of road surfaces is constructed based on the proposed spatiotemporal data association and the generative road surface models of covisibility regions. Finally, the proposed regionwise adaptive ROSE-constraint is integrated into the typical SLAM framework and tested on various custom and public datasets covering diverse terrain scenarios. Experimental results demonstrate the effectiveness and superiority of the proposed method compared to existing popular and state-of-the-art LiDAR-based SLAM solutions.

Current perception models in autonomous driving heavily rely on large-scale labelled 3D data, which is both costly and time-consuming to annotate. This work proposes a solution to reduce the dependence on labelled 3D training data by leveraging pre-training on large-scale unlabeled outdoor LiDAR point clouds using masked autoencoders (MAE). While existing masked point autoencoding methods mainly focus on small-scale indoor point clouds or pillar-based large-scale outdoor LiDAR data, our approach introduces a new self-supervised masked occupancy pre-training method called Occupancy-MAE, specifically designed for voxel-based large-scale outdoor LiDAR point clouds. Occupancy-MAE takes advantage of the gradually sparse voxel occupancy structure of outdoor LiDAR point clouds and incorporates a range-aware random masking strategy and a pretext task of occupancy prediction. By randomly masking voxels based on their distance to the LiDAR and predicting the masked occupancy structure of the entire 3D surrounding scene, Occupancy-MAE encourages the extraction of high-level semantic information to reconstruct the masked voxel using only a small number of visible voxels. Extensive experiments demonstrate the effectiveness of Occupancy-MAE across several downstream tasks. For 3D object detection, Occupancy-MAE reduces the labelled data required for car detection on the KITTI dataset by half and improves small object detection by approximately 2% in AP on the Waymo dataset. For 3D semantic segmentation, Occupancy-MAE outperforms training from scratch by around 2% in mIoU. For multi-object tracking, Occupancy-MAE enhances training from scratch by approximately 1% in terms of AMOTA and AMOTP.

The localization of autonomous vehicles mainly depends on GPS-based methods in large-scale outdoor scenarios. Unfortunately, GPS is vulnerable and not reliable in some regions, such as urban canyons. Odometry is taken as a complementary method for positioning in GPS-limited areas, but it suffers from accumulated errors, which limits its application. Topological map is a concise and low-cost representation of real world environment, which can be deemed as one of the information sources for localization. To realize accurate localization in the GPS-limited and large-scale scenes for autonomous vehicles, a generalized localization framework based on the topological map and odometry is proposed in this article. The odometry trajectory can be automatically aligned with the correct paths of the map through matching without the initial global position. Under this framework, odometry is deemed as the kinematic model while topological map is regarded as a global sensor that is utilized to correct the drift errors of odometry. As an example, the integration of visual odometry and topological map is used to validate the effectiveness of the proposed method on KITTI benchmark. Experimental results demonstrate that the proposed method can effectively constrain the drift errors of odometry, and it achieves state-of-the-art localization performance compared with existing map-aided methods.

LiNi0.5Mn1.5O4(LNMO) with a spinel structure is one of the most promising cathode materials choices for Li-ion batteries (LIBs). However, at a high operating voltages, the decomposition of organic electrolytes and the dissolution of transition metals, especially Mn(II) ions, cause unsatisfactory cycle stability. The initial application of a sodium alginate (SA)-xylan biopolymer as an aqueous binder aims to address the aforementioned problems. The SX28-LNMO electrode has a sizable discharge capacity, exceptional rate capability, and long-term cyclability with a capacity retention of 99.8% after 450 cycles at 1Cand a remarkable rate capability of 121 mAh g–1even at 10C. A more thorough investigation illustrated that SX28 binder provides a substantial adhesion property and generates a uniform (CEI) layer on the LNMO surface, suppressing electrolytes’ oxidative decomposition upon cycling and improving LIB performances. This work highlights the potential of hemicellulose as an aqueous binder for 5.0 V high-voltage cathodes.

Negative obstacle detection has been a challenging topic. In the previous researches, the distance that negative obstacles can be detected is so near that vehicles have to travel at a very low speed. In this paper, a negative obstacle detection algorithm from image sequences is proposed. When negative obstacles are far from the vehicle, color appearance models are used as the cues of detecting negative obstacles, while negative obstacles get closer, geometrical cues are extracted from stereo vision. Furthermore, different cues are combined in a Bayesian framework to detect obstacles in image sequences. Massive experiments show that the proposed negative obstacle detection algorithm is quite effective. The alarming distance for 0.8 m width negative obstacle is 18m, and the confirming distance is 10 m. This supplies more space for vehicles to slow down and avoid obstacles. Then, the security of the UGV running in the field can be improved remarkably.