3D object detection is critical for autonomous driving, yet it remains fundamentally challenging to simultaneously maximize computational efficiency and capture long-range spatial dependencies.We observed that Mamba-based models, with their linear state-space design, capture long-range dependencies at lower cost, offering a promising balance between efficiency and accuracy.However, existing methods rely on axis-aligned scanning within a fixed window, inevitably discarding spatial information. To address this problem, we propose WinMamba, a novel Mamba-based 3D feature-encoding backbone composed of stacked WinMamba blocks. To enhance the backbone with robust multi-scale representation, the WinMamba block incorporates a window-scale-adaptive module that compensates voxel features across varying resolutions during sampling. Meanwhile, to obtain rich contextual cues within the linear state space, we equip the WinMamba layer with a learnable positional encoding and a window-shift strategy.Extensive experiments on the KITTI and Waymo datasets demonstrate that WinMamba significantly outperforms the baseline. Ablation studies further validate the individual contributions of the WSF and AWF modules in improving detection accuracy. The code will be made publicly available.

The widespread adoption of 5G wireless communication devices has significantly increased the demand for precise 5G-based positioning services. This study presents a beam-switching time-of-arrival (TOA) ranging technique that leverages commercial 5G new radio (NR) signals and channel state information (CSI) extracted from the physical broadcast channel (PBCH). Initially, theoretical conditions for consistent multi-beam TOA estimation are derived, along with a multi-beam demodulation method designed to meet these conditions. To overcome the challenge of unknown base station (BS) radiation patterns, a signal quality-based beam-switching strategy is developed. Furthermore, a TOA tracking framework is introduced, integrating orthogonal matching pursuit (OMP) for multipath resolution, second-order frequency-locked loop (FLL)-assisted third-order delay-locked loop (DLL) for robust TOA tracking, and total variation regularization for anomaly removal. A software-defined radio (SDR) 5G receiver tailored for commercial beamforming BSs is implemented to validate the proposed system. With clock effects removed, the proposed system yields root-mean-square ranging errors of 3.73m and 4.99m in two distinct complex scenarios, corresponding to an average improvement of 69% in ranging accuracy over single-beam methods.

The proliferation of 5G wireless devices creates a pressing need for high-precision positioning services. This paper presents an integrated acquisition–and–tracking framework that leverages physical broadcast channel (PBCH) transmissions from commercial 5G New Radio (NR) base stations (BSs). To select the strongest downlink beam without prior knowledge of the BS radiation pattern, we use an adaptive beam-switching strategy based on reference signal received power (RSRP), reference signal received quality (RSRQ), and signal-to-noise ratio (SNR). For coarse acquisition, direction-of-arrival (DOA) and time-of-arrival (TOA) estimates are produced by a two-stage procedure that combines three-dimensional sparse Bayesian learning (SBL) with gradient-ascent off-grid refinement. A closed-loop tracker then continuously refines these estimates through azimuth-locked loop (ALL), elevation-locked loop (ELL), and delay-locked loop (DLL) modules. We further derive information-theoretic limits that bound acquisition and tracking and elucidate the key factors shaping these limits. The complete framework is implemented on a software-defined radio (SDR) 5G receiver and validated in outdoor field trials. Results indicate robust performance, with an average TOA root-mean-square error (RMSE) of 3.58 m, an azimuth RMSE of 5.77°, and an elevation RMSE of 2.74°, while requiring only 2.64 switching events per minute.

3D object detection is an indispensable component of autonomous driving. Due to the expensive and labor-intensive annotation required for full supervision, sparsely supervised 3D object detection is emerging as a promising alternative. Although some existing sparse supervision methods have achieved encouraging detection results, they do not perform well with distant or occluded objects. To address this issue, we propose a Consistency Regularization method for Sparsely-supervised 3D object detection(CRS3D). CRS3D consists of three modules: the Point Cloud Adjust module and the Consistency Loss module, which enhance the model’s ability to perceive distant objects by aligning predictions between the original and perturbed point clouds; and the Priori Ratio module, which optimizes the perception of occluded objects by imposing constraints based on priori information. In the KITTI benchmark, CRS3D achieves 87.4% 3D-mAP for the car class at easy difficulty using only 2% of the annotations, surpassing the accuracy of fully supervised methods.

Due to the sparsity and occlusion of point clouds, pure point cloud detection has limited effectiveness in detecting such samples. Researchers have been actively exploring the fusion of multimodal data, attempting to address the bottleneck issue based on LiDAR. In particular, virtual points, generated through depth completion from front-view RGB image, offer the potential for better integration with point clouds. Nevertheless, recent approaches fuse these two modalities in the region of interest (RoI), which limits the fusion effectiveness due to the inaccurate RoI region issue in the point cloud’s branch, especially in hard samples. To overcome it and unleash the potential of virtual points, while combining late fusion, we present leverage virtual point (LVP), a high-performance 3-D object detector which LVPs in early fusion to enhance the quality of RoI generation. LVP consists of three early fusion modules: virtual points painting (VPP), virtual points auxiliary (VPA), and virtual points completion (VPC) to achieve point-level fusion and global-level fusion. The integration of these modules effectively improves occlusion handling and improves the detection of distant small objects. In the KITTI benchmark, LVP achieves 85.45% 3-D mAP. As for large dataset nuScenes, we could improve the detection accuracy of large objects by compensating for errors in depth estimation. Without whistles and bells, these results establish LVP as an impressive solution for a 3-D outdoor object detection algorithm.

When the number of available training views is limited, NeRF and 3DGS will soon overfit the optimization and learn the wrong scene geometry. For this challenge, a common solution is to provide depth prior as supervision to correct scene geometry. In this work, we present Geometric Regularized 3D Gaussian Splatting (GeoRGS), a priors-independent method for improving novel view synthesis from sparse inputs. We analyze the problems of the density control strategy in 3DGS with sparse inputs, and find that correcting the erroneous Gaussian growth trend at the beginning of training is effective in mitigating overfitting. Based on this analysis, we propose two geometric regularization methods that do not require prior information. One is based on selecting seed patches of 3D Gaussian from the scene, which guides growth to form correct scene geometry, while the other focuses on regularizing depth similarity between object surfaces and edges. GeoRGS achieves state-of-the-art performance in novel view synthesis from sparse input on LLFF, Blender, RealEstate10K and MipNeRF360 datasets, while also demonstrating significantly faster training speeds and rendering efficiency compared to other baselines.

Location-based service in the indoor environment is playing a crucial role in different application scenarios. The introduction of technologies, such as ultradense network and massive multiple-input multiple-output enables fifth-generation (5G) cellular signals, as a new generation of cellular network signals, to show unique advantages in indoor positioning. This article describes 5G reference signal structures that can be used for navigation. A high-precision time-of-arrival estimation method based on 5G downlink signal is proposed that can be realized by edge computing. A software-defined receiver (SDR) based on machine learning to extract navigation observations from 5G signals is then developed. In simulation, the error sources of SDR in additive white gaussian noise channel and multipath channel were analyzed, and the possible ranging accuracy achieved by 5G signals in the developed SDR was evaluated. In field experiments, commercial 5G signals deployed by operators were collected, and the performance of SDR in practical applications was evaluated. The feasibility in practical applications of the proposed SDR is demonstrated, and high pseudorange measurement accuracy can be achieved.

The fifth-generation (5G) networks have been massively deployed in commerce. The new features introduced by 5G networks are beneficial to wireless positioning. In this study, the performance of indoor positioning with commercial 5G new radio (NR) signals is investigated, and the channel state information (CSI) extracted from the downlink synchronization signal block is utilized. Considering the limited 5G NR base station (known as gNodeB) is hearable indoors, the fingerprint method is used, and an indoor positioning system termed iPos-5G is developed. The system consists of four components. First, a module of quality control is applied for CSI preprocessing. Second, an unsupervised deep-autoencoder network is utilized to reconstruct CSI features. Third, by supervised learning, a radial basis function is improved to optimize the probability model for similarity calculations. Finally, an amplitude-phase probability fusion function is proposed for positioning by weighting the coordinates of reference points. To verify the effectiveness of iPos-5G, indoor field tests are carried out in the scenarios of an office and a corridor. The test results show that iPos-5G achieves mean absolute errors of 2.14 and 2.81 m and standard deviation of the errors of 1.07 and 1.66 m, which outperforms the compared CSI fingerprint methods in terms of positioning accuracy and stability.

Selecting local features is crucial in generating robust compact descriptors for mobile visual search. The state-of-the-art MPEG Compact Descriptors for Visual Search (CDVS) standard has utilized the intrinsic characteristics (e.g., scale, orientation, peak, center distance, etc.) of interest points to select salient local features for selective aggregation and compression of local feature descriptors at different bit rates. In particular, the statistics of center distance was considered as an important attribute to select features in mobile visual search, which heavily relies on the assumption of a centralized object in a 2-dimensional query image. However, the ad-hoc assumption would probably fail to delineate query objects in a cluttered scene. In this paper, we propose to incorporate the depth cue to select local features. As most mobile phones are not yet equipped with depth sensor, we recover the disparity of local features through an auxiliary image to fast estimate the depth of a query image. The experiments have shown that, the incorporation of depth cue into feature selection can significantly improve the retrieval performance of the state-of-the-art CDVS compact descriptors at lower bit rates. For example, the mAP is improved from 84.5% to 88.6% at 512 bytes.