Bernard Mulgrew

The efficient extraction of useful information from radio frequency (RF) sensors is one important application for artificial intelligence (AI) and machine learning (ML) approaches. In particular, there is a desire to maximise efficiency when sharing positioning, navigation, and timing (PNT) information captured by distributed networks of low size, weight, power, and cost (SWAP-C) RF sensors when operating in congested or contested electromagnetic environments (EMEs). By implementing effective PNT information sharing strategies, these networks can more easily position the sensors or characterise targets of interest. In this work, we propose a novel ML-inspired compression design framework that improves efficiency when sharing PNT information in a network of sensors receiving radar waveforms. In addition, through novel learning procedures, the network can adapt to unforeseen EMEs such that network efficiency can be maintained in the presence of unforeseen RF waveforms and sensor surroundings. We show that our intelligent, model-driven, ML-inspired data reduction strategies can outperform alternative strategies that do not best-utilise the information content of waveforms within the EME. In addition, we demonstrate the ability of our strategies to adapt to changing mission goals by balancing different types of PNT information and learning from developing EMEs.

We consider the problem of detecting small and manoeuvring objects with staring array radars. Coherent processing and long-time integration are key to addressing the undesirably low signal-to-noise/background conditions in this scenario and are complicated by the object manoeuvres. We propose a Bayesian solution that builds upon a Bernoulli state space model equipped with the likelihood of the radar data cubes through the radar ambiguity function. Likelihood evaluation in this model corresponds to coherent long-time integration. The proposed processing scheme consists of Bernoulli filtering within expectation maximisation iterations that aims at approximately finding complex reflection coefficients. We demonstrate the efficacy of our approach in a simulation example.

We consider the problem of detecting small and manoeuvring objects with staring array radars. Coherent processing and long-time integration are key to addressing the undesirably low signal-to-noise/background conditions in this scenario and are complicated by the object manoeuvres. We propose a Bayesian solution that builds upon a Bernoulli state space model equipped with the likelihood of the radar data cubes through the radar ambiguity function. Likelihood evaluation in this model corresponds to coherent long-time integration. The proposed processing scheme consists of Bernoulli filtering within expectation maximisation iterations that aims at approximately finding complex reflection coefficients. We demonstrate the efficacy of our approach in a simulation example.

In this paper, an active multi-receiver low-THz imaging radar is used to generate 3D images of automotive environments by combining time of arrival information with backprojection techniques to give height profiles of the objects in the scene. The large bandwidth of this system facilitates non-coherent height reconstruction using time of arrival alone. In doing so, the constraints on receiver coherency are relaxed compared to many height finding techniques. 3D reconstructions of common roadside targets and scenes are generated through a modified form of backprojection, adapted to use the time of arrival information of the signals from multiple receivers with different sensitivities, which are arranged in a quasi-multi-static configuration. Here it is experimentally shown that the use of three vertically aligned receivers is sufficient to provide a fine resolution in range, cross-range, and height. The system is used to generate 3D reconstructions of a number of different scenarios, including pedestrians moving through a scene, and complex real-world road scenes.

In this article, we present an adaptive subaperture integration method for wide-angle synthetic aperture radar (SAR) for improved imaging, with emphasis on short-to-medium range applications. In order to avoid full-aperture integration, traditional approaches use fixed-width subapertures, which may not conform to the persistence angle of the scatterers. Coherent integration gains over the aperture are possible if integration is carried out over the persistence angle of the scatterers, because integrating shorter than the persistence angle may spread the scattering response across multiple subapertures or, conversely, integrating more than the persistence angle may cause noise accumulation along with the useful signal. In this article, we propose to use change-point detection methods to estimate the persistence widths of the scatterers, and consequently enhance the coherent integration gains, resulting in improved imaging. We compare our proposed methods with the standard integration approaches as well as a recently proposed adaptive integration approach. We provide qualitative and quantitative analyses to prove that our proposed methods outperform the existing approaches. We present experimental results on the real-data of our low-terahertz radar as well as a publicly available dataset to validate our claims.

Receiver nonlinearities pose a serious risk to the functionality of modern radars as they can compromise the sensor’s immunity to interfering signals. With the radio frequency (RF) spectrum becoming increasingly crowded, it is now more important than ever that the sensor can maintain system performance when exposed to mutual interference. In this article, we present a nonlinear compressive sensing (NCS) solution, which, unlike the standard nonlinear equalization (NLEQ) techniques, is designed around the forward nonlinearity rather than the inverse. Importantly, in this article, the NCS theory is extended to include nonlinear memory. Furthermore, a radar-specific formalization is derived, which allows the nonlinear optimization to exploit the unique properties of pulsed-Doppler radar processing. As a result, the NCS solution can successfully restore system sensitivity back to the linear case when in-band interference drives the radar receiver into its nonlinear regime. Additionally, it is shown that the technique can consistently mitigate complex nonlinear memory effects generated in the RF receiver. This is a significant result as it proves that forward modeling techniques are a viable alternative to NLEQ. This is of particular importance to radar systems as they provide a far more explicit formalization to mitigate nonlinear memory effects.

In this article, we introduce a frequency-based waveform design to maximize binary target classification. The optimization problems can be solved via optimal solvers available in the literature. The presented formulation is applicable to the classification scenarios where the extended target frequency responses (TFRs) are complex, random, and normally distributed with unequal mean vectors. However, their covariance matrices are either identical or different. For the former, i.e., identical covariance matrices, the optimization problem consists of a cost function that has been obtained by deriving a closed-form expression of the probability of misclassification. For the latter, i.e., different covariance matrices, we present an optimization problem where the cost function has been obtained via the Fisher separation function. For this, we also present a closed-form expression for an approximate solution to the optimization problem. We show that the proposed solution achieves performance levels comparable to the exact solution of the optimization problem obtained via state-of-the-art optimization solvers while incurring low computational complexity. We expand on the two main scenarios by introducing clutter (i.e., signal-dependent interference) in the signal model and studying possible closed-form designs in extreme waveform energy levels. Simulations are conducted using synthetically generated data in addition to the civilian vehicle data from the Moving and Stationary Target Acquisition and Recognition (MSTAR) dataset, in order to validate the proposed method.

In this article, we propose a novel concept of cross-learning in order to improve synthetic aperture radar (SAR) images by learning from the camera images, in the manifold domain. We present multilevel abstraction approaches to materialize knowledge transfer between these two very different modalities (i.e., the radar and the camera), namely, a canonical correlation analysis-based approach and a manifold alignment-based approach. We provide experimental results on real data, along with qualitative as well as quantitative analyses, to validate the proposed methodologies.