Research

Generalizable machine learning methods for wireless propagation modeling

Machine learning (ML) methods offer an alternative route to formulating propagation models that can potentially combine the efficiency of empirical models with the accuracy of physics-based models. These methods are based on training a model, such as an artificial neural network, with measured and/or simulated data corresponding to various quantities of interest, such as path loss and received signal strength.

A fundamental challenge for such models is their ability to “generalize”: to rapidly and accurately solve problems with specifications that are beyond those included in the training set. To this end, we presented ML models for indoor and tunnel propagation that can generalize with respect to the geometry, position of transmitter and receiver and operating frequency, in (Seretis and Sarris, “Towards Physics-Based Generalizable Convolutional Neural Network Models for Indoor Propagation”, to appear on IEEE Trans. on Antennas Propagat., 2022).

Computational models for electromagnetic wave propagation in wireless communication systems

We have contributed ray-tracing, vector parabolic equation (VPE) and hybrid methods for the modeling of electromagnetic wave propagation in wireless communication channels.

In (Zhang and Sarris, IEEE Trans. on Antennas Propagat., vol. 64 (3), 2017), we extended the VPE method to embed antennas of arbitrary (realistic) patterns and efficiently model their radiation within tunnels. In (Zhang and Sarris, IEEE Trans. on Antennas Propagat., vol. 67 (4,) 2019), we presented a rigorous, VPE-based approach for the stochastic modeling of wall surface roughness and its impact on wave propagation in tunnels and mines. Moreover, we presented a hybrid technique to model diffuse scattering in millimeter-wave communication systems with ray-tracing in (Bakirtzis, Hashimoto and Sarris, IEEE Trans. on Antennas Propagat., vol. 69 (6), 2020).

Integration of physics-based models in communication system analysis and design

Physics-based wireless propagation modeling and network protocol design have evolved over decades as orthogonal areas in communication systems research. This fragmented approach does not exploit available efficiencies when planning and deploying communication systems. On the other hand, the integration of these two areas can pave the way for far better, more secure and more reliable wireless systems within the increasingly complex landscape of existing and emerging wireless services.

In (Sood et al., IEEE Trans. on Antennas Propagat., 66 (12), 2018) and (Sood et al., IET Microwaves, Antennas Propagat., 13 (8), 2019) we demonstrated robust wireless network protocol analysis and design, harnessing the power of accurate models of electromagnetic propagation in wireless systems for train control. The anticipated benefits of this research extend to any safety-critical communication system, for commercial and defence applications.

Uncertainty quantification in numerical electromagnetics

Research on computational electromagnetics has focused on the accurate modeling of arbitrarily complex yet well-defined structures. However, micro- and nano-fabrication tolerances often play an important role in the performance of electromagnetic and optical structures.

We have contributed efficient methods for accurately evaluating geometric uncertainties in electromagnetic geometries over broad frequency bandwidths: with the polynomial chaos expansion in (Austin and Sarris, IEEE Trans. on Microwave Theory Tech., 61 (12), 2013); overcoming round-off errors to compute electromagnetic field derivatives of any degree with respect to geometric and material parameters to machine precision in (Liu and Sarris, IEEE Trans. on Antennas Propagat., 67 (6), 2019) employing the complex-step derivative approximation, and in (Liu and Sarris, IEEE Trans. on Microwave Theory Tech., 67 (10), 2019).

Convex and semi-definite relaxation optimization of wireless power transfer systems and antennas

We pioneered a semi-analytical, intuitive mathematical framework for the optimization of multi-transmitter wireless power transfer systems in (Lang, Ludwig and Sarris, IEEE Trans. on Antennas Propagat., 65 (10), 2017). We extended this work to systems with passive couplers and metasurfaces in (Lang and Sarris, IEEE Trans. on Antennas Propagat., 62 (9), 2014). To that end, we employed convex optimization, while non-convex constraints (such as the physically intuitive constraint that input powers to all transmitter ports should be positive) were included via a semi-definite relaxation approach, introduced in (Lang and Sarris, IEEE Trans. on Microwave Theory Tech., 65 (11), 2017).