arXiv subjects

Salman Liaquat

Publications and source records attributed to Salman Liaquat.

8 recordsLinked to original sources

arXiv2026

Design and Experimental Validation of a Multiband Cross-Polarization Conversion (CPC) Metasurface for Radar Cross Section (RCS) Reduction

Radar cross-section (RCS) reduction is a fundamental requirement in modern stealth technology, playing a critical role in the low-observable performance of aerial and naval platforms. Among the various passive RCS reduction strategies, including radar-absorbing materials, absorptive coatings, and artificially engineered surfaces, metasurface-based cross-polarization conversion has emerged as a compelling approach owing to its structural simplicity and low profile. In this work, a single-layer cross-polarization conversion (CPC) metasurface developed on a cost-effective FR4 dielectric substrate (relative permittivity 4.4, loss tangent 0.02) is proposed for multiband RCS reduction. The designed structure achieves a polarization conversion ratio (PCR) exceeding 95% at three distinct operating frequencies of 7.8 GHz, 11.7 GHz, and 18 GHz, spanning the C-, X-, and Ku-bands, which directly translates into a monostatic RCS reduction exceeding 10 dBsm at the corresponding bands. The metasurface further demonstrates stable polarization conversion performance under oblique incidence up to 60 degrees, confirming its suitability for wide-angle illumination conditions encountered in practical deployment scenarios. Experimental validation conducted in an anechoic chamber confirms close agreement with full-wave electromagnetic simulations, substantiating the reliability of the fabricated prototype. The proposed design offers a lightweight, low-cost, and high-performance candidate for multiband stealth and low-observable platform applications.

eess.SP↗

arXiv2026

An X-Band Monopulse Direction-Finding Receiver Based on a Rat-Race Comparator and a $2\times2$ Antipodal Vivaldi Array

This paper presents the design, fabrication, and experimental validation of a compact monopulse direction-finding (DF) receiver operating in the X-band. The receiver combines a $2 \times 2$ antipodal Vivaldi antenna array with a rat-race (180-degree hybrid) comparator network that simultaneously synthesizes the sum ($Σ$), azimuth-difference ($Δ_{az}$), and elevation-difference ($Δ_{el}$) channels required for monopulse processing, removing the need for mechanical scanning. The Vivaldi element provides good impedance matching across 8-12 GHz with a simulated realized gain of approximately 6.44 dBi, and the comparator exhibits the 180-degree phase balance at its difference ports required for angle encoding, with the fabricated prototype preserving the matched, balanced response measured across the band. Three LTC5564 envelope detectors convert the channel outputs to DC voltages that are digitized by an on-board microcontroller and processed in software to form the monopulse ratios and estimate the direction of arrival (DoA), with the result displayed in real time on a MATLAB graphical user interface. Azimuth direction finding is experimentally demonstrated over fourteen angles spanning -20 to 19 degrees at 11 GHz, yielding a root-mean-square error (RMSE) of 7 degrees; the architecture is directly extensible to elevation. The result is a passive, low-cost, and fully integrated receiver suitable for radar sensing, electronic warfare, and surveillance.

eess.SP↗

arXiv2026

A Closed-Form Noise-Sensitivity Asymmetry for Causal Branch Selection in Minimal-Array TDoA Localization

Minimal-array time-difference-of-arrival (TDoA) localization with three planar receivers reduces to a scalar quadratic whose two roots can both be feasible target positions, leaving a branch-selection ambiguity intrinsic to the minimal geometry. Because a minimal array is the cheapest, most deployable passive configuration, this ambiguity is conventionally broken only by adding a fourth receiver or an angle sensor, sacrificing the very minimality that motivates the array. This article resolves it from the measurements alone. Implicit differentiation shows that the two roots share an identical sensitivity denominator, so classical root conditioning cannot separate them and the entire asymmetry resides in the numerator. A closed-form analysis then yields an exact sign condition for the asymmetry, and over the feasible interior the more sensitive root is the outer, physical one, away from two algebraic degeneracy loci. A causal, constant-memory selector smooths a per-root variability statistic, whose expectation is shown to be proportional to the per-root sensitivity in the noise-dominated regime, and selects the larger one. Across a dimensionless receiver atlas the physical root is the more sensitive at a median of 85\% of two-feasible-root operating points (median sensitivity ratio 16), and, because the more sensitive root is the outer one, a simple outer-root rule attains near-0.99 branch-selection accuracy on this interior, matched online by a causal, constant-memory realization. The decisive empirical finding is that smoothness- and continuity-based disambiguation, the natural alternatives, invert under timing noise and fall below chance.

eess.SP↗

arXiv2026

$κ$: A Geometry-Quality Metric Complementary to GDoP for Closed-Form TDoA Multilateration

The Geometric Dilution of Precision (GDoP) characterizes the noise sensitivity of a Time-Difference-of-Arrival (TDoA) localization system, but does not capture every way the analytical multilateration solution can become ill-conditioned. We introduce a complementary geometry-quality metric $κ$, the leading coefficient of the closed-form TDoA solver's quadratic, and derive its $N$-dimensional generalization through a vectorized formulation. Two closed-form algebraic identities relate $κ$ to the Jacobian determinant of the measurement model and to the quadratic's discriminant, establishing that the system exhibits exactly two distinct singularity loci: branch divergence and the Jacobian/branch-merge locus flagged by GDoP. A Cramér--Rao-bound-linked closed form for the noise sensitivity $σ_κ$ under the standard Gaussian ToA model is validated against Monte Carlo to 2% median relative error. An empirical atlas over a dimensionless geometry parameter space confirms both identities at machine precision and shows that $κ$-bad regions and GDoP-bad regions are non-trivially disjoint in target space, establishing the two metrics as genuinely complementary. A case study on a four-node operational array, with per-sensor time of arrival (ToA) noise estimated empirically from Automatic Dependent Surveillance Broadcast (ADS-B)-paired over-the-air captures, shows that the theory-predicted threshold and a Monte-Carlo-measured operational threshold agree on the per-subsystem ordering at the deployment noise level. Their ratio is approximately constant across the three two-dimensional subsystems, serving as a deployment-specific calibration constant between the algebraic $κ$-noise floor and the downstream operational threshold, analogous in spirit to the standard relation linking GDoP to the circular error probable.

eess.SP↗

arXiv2026

Feature-Level Robustness of Physics-Guided Micro-Doppler Descriptors for classification of Drones and Birds

Micro-Doppler signatures are a proven modality for discriminating between drones and birds, but their reliability degrades in low-SNR, data-constrained settings where deep learning models often fail. This paper presents a systematic study of ten statistical and physics-motivated handcrafted features for micro-Doppler classification under controlled signal degradation, using a publicly available 77 GHz FMCW radar dataset. Spectrograms are corrupted with additive white Gaussian noise, phase noise, and their combination across SNRs from -10 dB to 10 dB and phase noise levels from 1 to 10 degrees. Features are evaluated using stratified 5-fold cross-validation with Support Vector Machine and Random Forest classifiers, using fixed hyperparameters across all noise conditions. On clean data, both models achieve mean accuracy of 0.916, with F1 scores of 0.909 (SVM) and 0.892 (Random Forest). Under severe noise, entropy-based and side-lobe features remain robust, yielding F1 scores up to 0.773 and 0.831, respectively. Permutation-based importance analysis shows that some features retain complementary discriminative power even when their individual importance is low. These results highlight the value of principled feature design and provide insight into feature robustness for interpretable radar classification systems.

eess.SP↗

arXiv2025

Dual RIS-Assisted Monostatic L-Band Radar Target Detection in NLoS Scenarios

The use of a single Reconfigurable Intelligent Surface (RIS) to boost the signal-to-noise ratio (SNR) at the radar offers significant improvement in detecting targets, especially in non-line-of-sight (NLoS) scenarios. However, there are scenarios where no path exists between the radar and the target, even with a single RIS-assisted radar, due to other present obstacles. This paper derives an expression for SNR in target detection scenarios where dual RISs assist a monostatic radar in NLoS situations. We calculate the power received at the radar through a dual RIS configuration. We show that the SNR performance of RIS-assisted radars can improve with known locations of the radar and RISs. Our results demonstrate that the required accuracy in target localization can be achieved by controlling the number of RISs, the number of unit cells in each RIS, and properly selecting the locations of RISs to cover the desired region. The performance of dual RIS-assisted radar systems can surpass that of single RIS-assisted radar systems under favourable alignment and sufficiently large RIS sizes.

eess.SP↗

arXiv2025

Using Continual Learning for Real-Time Detection of Vulnerable Road Users in Complex Traffic Scenarios

Pedestrians and bicyclists are among the vulnerable road users (VRUs) that are inherently exposed to intricate traffic scenarios, which puts them at increased risk of sustaining injuries or facing fatal outcomes. This study presents an intelligent adaptive system that uses the YOLOv8-Dynamic (YOLOv8-D) algorithm that detects vulnerable road users and adapts in real time to prevent accidents before they occur. We select YOLOv8x as the detector by comparing it with other state-of-the-art object detection models, including Faster-RCNN, YOLOv5, YOLOv7, and variants. Compared to YOLOv5x, YOLOv8x shows improvements of 12.14% in F1 score and 45.61% in mean Average Precision (mAP). Against YOLOv7x, the improvements are 21.26% in F1 score and 128.44% in mAP. Our algorithm integrates continual learning ability in the architecture of the YOLOv8 detector to adjust to evolving road conditions flexibly, ensuring adaptability across multiple dataset domains and facilitating continuous enhancement of detection and tracking accuracy for VRUs, embracing the dynamic nature of real-world environments. In our proposed framework, we optimized the gradient descent mechanism of YOLOv8 model and train our optimized algorithm on two statistically different datasets in terms of image viewpoint and number of classes to achieve a 21.08% improvement in F1 score and a 31.86% improvement in mAP as compared to a custom YOLOv8 framework trained on a new dataset, thus overcoming the issue of catastrophic forgetting, which occurs when deep models are trained on statistically different types of datasets.

eess.IV↗

arXiv2024

SDR-Based Metal Classification using Spectrogram Images from Micro-Doppler Signatures

Metallic materials such as brass, copper, and aluminum are used in numerous applications, including industrial manufacturing. The vibration characteristics of these objects are unique and can be used to identify these objects from a distance. This research presents a methodology for detecting and classifying these metallic objects using the vibration dynamics induced by their micro-Doppler signatures. The proposed approach utilizes image processing techniques to extract pivotal features from spectrograms. These spectrograms originate from micro-Doppler signatures of data collected during controlled laboratory experiments where signals were transmitted towards vibrating metal sheets, and the ensuing reflections were recorded using a software-defined radio (SDR). The spectrogram data was augmented using geometric transformation to train a convolutional neural network (CNN) based machine learning model for object classification. The results indicate that the proposed CNN model achieved an accuracy of more than 95% in classifying metals into brass, copper, and aluminum. This research could be used to understand the foundations of classifying spectrogram images using micro-Doppler signatures for its applications towards enhancing the sensing capabilities in industrial and defense applications.

eess.SP↗