Lei Huang

Nowadays lightweight radar has become an important technique in autonomous driving systems for its all-weather sensing ability. Synthetic aperture radar (SAR) imaging based on one-bit sampling is able to considerably simplify system deployment and cut down manufacturing costs. However, a high oversampling rate is usually required to avoid the performance degradation resulting from aliasing of one-bit quantization noise. This will counteract the storage saving advantages brought by low quantization level. In this paper, to facilitate its applicability on vehicle platforms, a lightweight strategy of 1.5-bit SAR imaging is devised to inherit the merit of one-bit quantization in simplification but circumvent the requirement of high sampling rate. Compared with one-bit quantization, only additional 0.5 bit per sample is introduced to alleviate the awkward harmonic problem due to the non-linear effect of quantization. In this way, the quality of SAR imagery can be significantly improved. Besides, the enhanced performance of the proposed method does not rely on the sampling rate. This will benefit the whole procedure of imaging and lighten the system burden and cost consequently. Theoretical analysis and experimental results from a real vehicular SAR system showcase the superiority of our 1.5-bit quantization approach.

Constrained by the physical geometry of microphone arrays, acoustic beamforming faces limitations in its operational frequency range and scanning scale. Non-synchronous measurements (NSM) offer a viable solution by moving an array to construct a virtual large dense array. Our team, in a previous study , proposed a tensor completion method based on the Tensor Nuclear Norm (TNN), which addresses the frequency selection problem by reconstructing the Cross-Spectral Tensor (CST) of broadband signals . However, the TNN , as a convex surrogate for tensor rank, has a limitation: it assigns equal weight to all singular values, which restricts its performance under low signal-to-noise ratio (SNR) conditions. As an extension and improvement of that work , this paper proposes a novel approach: a non-synchronous measurements method for broadband multiple-sound-source localization based on Laplacian-enhanced low-rank tensor completion. This method utilizes the Laplace function as a superior non-convex surrogate, enabling it to automatically assign weights based on singular value importance, thereby achieving a more robust recovery of the low-rank tensor structure. Simulations and experimental studies demonstrate that, compared to TNN and other methods, the proposed L-TNN algorithm produces sound maps with lower side-lobes and higher accuracy, especially under challenging conditions of low SNR and sparse measurements.

This paper addresses the jamming detection problem in colored noise environments. Low-power attacks are particularly dangerous due to their stealthiness and persistent harmful effects. The fluctuations caused by colored noise backgrounds further exacerbate this issue, making it easier for these attacks to remain concealed within the noise. We apply a source enumerator, namely the second-order-difference (SOD) algorithm that is robust against both white and colored noise, for jamming detection, and the resulting method is referred to as the SOD jamming detection (SODJD) method. Our analysis of jamming power’s impact on SODJD performance demonstrates that while SODJD can address colored noise, it is limited in scenarios with low-power signals mixed with several higher-power signals. To overcome these limitations, we propose an invariant SOD jamming detection (ISODJD) method by redesigning the criterion rule and introducing an adjustable parameter. Further analysis reveals that larger parameter values lead to improved detection performance, and the increased false alarm rate can be mitigated in massive channel systems, making ISODJD suitable for cooperative wireless communication scenarios. We implement our schemes and conduct extensive performance comparisons. Simulations demonstrate the effectiveness and superiority of the proposed method. Compared to the SODJD method, a 60-antenna MIMO system achieves a 4 dB improvement in detection sensitivity for low-power jamming signals under colored noise, aligning with our theoretical analysis.

This paper addresses the multistatic localization problem with the receiving sensors distributed on a rigid body (RB) whose position and orientation are not known. We propose to jointly estimate the unknown object position together with the RB position and orientation, simply called joint object and rigid body localization (JORBL), using time delay (TD) measurements from both the direct-propagation and indirect-reflection paths between the transmitters and sensors. Two scenarios are considered, one with synchronization between the sensors and transmitters and the other without. We begin by investigating the benefits of having the direct-path TD measurements compared to using the indirect-path TDs only for JORBL, in terms of estimation accuracy and minimum number of transmitters and sensors needed, through analyzing the Cramér-Rao lower bound (CRLB). Afterward, in the presence of synchronization, we propose a JORBL method by formulating a non-convex constrained weighted least squares problem and applying semidefinite relaxation to obtain the solution by semidefinite programming. The proposed method is then extended to the case without synchronization. The mean squared error analysis demonstrates that the proposed methods can achieve the CRLB accuracy for small Gaussian noise, which is further confirmed by numerical simulations.

This paper studies a beyond-diagonal reconfigurable intelligent surface (BD-RIS)-aided integrated sensing and communications (ISAC) system enhanced by rate-splitting multiple access (RSMA) to improve sensing and communication abilities with flexible interference management. Unlike conventional RIS, BD-RIS is a generalized RIS paradigm with a non-diagonal scattering matrix, offering additional degrees of freedom. Moreover, RSMA is a novel multiple access framework for effective and flexible interference control. In this work, we formulate a sensing signal-to-noise ratio maximization problem with reduced multi-user interference (MUI), which is challenging to jointly design the BD-RIS scattering, precoding, and rate splitting due to strong cross-domain coupling. To address these issue, we devise a two-stage optimization framework. First, the BD-RIS scattering matrix is optimized via symmetric unitary projection to improve the channel gain, where a closed-form solution is derived. Then, the weighted minimum mean square error framework is customized for precoding and rate splitting optimization. Simulation results validate the advantages of the proposed system and the algorithm, showing that the BD-RIS-aided RSMA-enhanced ISAC system is able to effectively improve the sensing performance and flexibly control the MUI.

The operational safety and efficiency of Intelligent Transportation Systems (ITS) critically depend on reliable environmental perception, which is severely compromised in adverse weather conditions where conventional vision sensors fail. Synthetic aperture radar (SAR) technology, which can deliver high-resolution images under severe weather and poor-light conditions, is being increasingly integrated into automotive systems. However, the challenges like dynamic trajectory estimation and real-time SAR imaging remain significant obstacles. This paper proposes a novel automotive SAR system designed to function as an all-weather perception enabler for autonomous driving. By employing sub-aperture scheme for SAR system, the approach eliminates the need for complicated and time-consuming range cell migration and motion correction. Due to the short coherent accumulation, the instant range doppler algorithm enables high-efficient SAR imagery generation while maintaining two-dimensional (2D) high-resolution performance, which allows precise target detection. The theoretical analysis and experimental results confirm the effectiveness of the proposed scheme. Then, a benchmark is firstly established, to advance the data-driven radar perception in ITS. To validate the radar data, diverse targets in SAR images are annotated, providing a robust database support for the automotive SAR target recognition. Furthermore, the value of this dataset for ITS is demonstrated by evaluating various mainstream deep learning methods on object detection tasks, confirming its potential to enhance the robustness of perception modules in intelligent vehicles.