The data-science revolution is poised to transform the way photonic systems are simulated and designed. Photonic systems are, in many ways, an ideal substrate for machine learning: the objective of much of computational electromagnetics is the capture of nonlinear relationships in high-dimensional spaces, which is the core strength of neural networks. Additionally, the mainstream availability of Maxwell solvers makes the training and evaluation of neural networks broadly accessible and tailorable to specific problems. In this Review, we show how deep neural networks, configured as discriminative networks, can learn from training sets and operate as high-speed surrogate electromagnetic solvers. We also examine how deep generative networks can learn geometric features in device distributions and even be configured to serve as robust global optimizers. Fundamental data-science concepts framed within the context of photonics are also discussed, including the network-training process, delineation of different network classes and architectures, and dimensionality reduction. Neural networks can capture nonlinear relationships in high-dimensional spaces and are powerful tools for photonic-system modelling. This Review discusses how deep neural networks can serve as surrogate electromagnetic solvers, inverse modelling tools and global device optimizers.

Deep neural networks for the evaluation and design of photonic devices

The data sciences revolution is poised to transform the way photonic systems are simulated and designed. Photonics are in many ways an ideal substrate for machine learning: the objective of much of computational electromagnetics is the capture of non-linear relationships in high dimensional spaces, which is the core strength of neural networks. Additionally, the mainstream availability of Maxwell solvers makes the training and evaluation of neural networks broadly accessible and tailorable to specific problems. In this Review, we will show how deep neural networks, configured as discriminative networks, can learn from training sets and operate as high-speed surrogate electromagnetic solvers. We will also examine how deep generative networks can learn geometric features in device distributions and even be configured to serve as robust global optimizers. Fundamental data sciences concepts framed within the context of photonics will also be discussed, including the network training process, delineation of different network classes and architectures, and dimensionality reduction.

A review of the most recent advances in deep learning (DL) as applied to electromagnetics (EM), antennas, and propagation is provided. It is aimed at giving the interested readers and practitioners in EM and related applicative fields some useful insights on the effectiveness and potentialities of deep neural networks (DNNs) as computational tools with unprecedented computational efficiency. The range of considered applications includes forward/inverse scattering, direction-of-arrival estimation, radar and remote sensing, and multi-input/multi-output systems. Appealing DNN-based solutions concerned with localization, human behavior monitoring, and EM compatibility are reported as well. Some final remarks are drawn along with the indications on future trends according to the authors’ viewpoint.

DNNs as Applied to Electromagnetics, Antennas, and Propagation—A Review

Unmanned aerial vehicle (UAV) networks are playing an important role in various areas due to their agility and versatility, which have attracted significant attentions from both the academia and industry in recent years. As an integration of embedded systems with communication devices, computation capabilities and control modules, the UAV network could build a closed loop from data perceiving, information exchanging, decision making to the final execution, which tightly integrates cyber processes into physical devices. Therefore, the UAV network could be considered as a cyber physical system (CPS). Revealing coupling effects among the three interacted CPS components, i.e., communication, computation and control, is envisioned as the key to properly utilize all available resources and hence improve the performance of the UAV network. In this paper, we present a comprehensive survey on UAV networks from a CPS perspective. Firstly, we review the basics and advances of the three CPS components in UAV networks. Then we look inside to investigate how these components contribute to the system performance by classifying UAV networks into three hierarchies, i.e., cell level, system level, and system of system level. Furthermore, the coupling effects among these CPS components are explicitly illustrated, which could be enlightening to deal with the challenges in each individual aspect. New research directions and open issues are discussed at the end of this survey. With this intensive literature review, we intend to provide a novel insight into the state of the art in UAV networks.

Survey on Unmanned Aerial Vehicle Networks: A Cyber Physical System Perspective

In recent years, metasurfaces have emerged as revolutionary tools to manipulate the behavior of light at the nanoscale. These devices consist of nanostructures defined within a single layer of metal or dielectric materials, and they offer unprecedented control over the optical properties of light, leading to previously unattainable applications in flat lenses, holographic imaging, polarimetry, and emission control, amongst others. The operation principles of metaoptics include complex light–matter interactions, often involving insidious near-field coupling effects that are far from being described by classical ray optics calculations, making advanced numerical modeling a requirement in the design process. In this contribution, recent optimization techniques used in the inverse design of high performance metasurfaces are reviewed. These methods rely on the iterative optimization of a Figure of Merit to produce a final device, leading to freeform layouts featuring complex and non-intuitive properties. The concepts in numerical inverse designs discussed herein will push this exciting field toward realistic and practical applications, ranging from laser wavefront engineering to innovative facial recognition and motion detection devices, including augmented reality retro-reflectors and related complex light field engineering.

Numerical optimization methods for metasurfaces

In this work, machine learning methods are applied to high-speed channel modeling for signal integrity analysis. Linear, support vector, and deep neural network (DNN) regressions are adopted to predict the eye-diagram metrics, taking advantage of the massive amounts of simulation data gathered from prior designs. The regression models, once successfully trained, can be used to predict the performance of high-speed channels based on various design parameters. The proposed learning-based approach saves complex circuit simulations and substantial domain knowledge altogether, in contrast to alternatives that exploit novel numerical techniques or advanced hardware to speed up traditional simulations for signal integrity analysis. Our numerical examples suggest that both support vector and DNN regressions are able to capture the nonlinearities imposed by transmitter and receiver models in high-speed channels. Overall, DNN regression is superior to support vector regression in predicting the eye-diagram metrics. Moreover, we also investigate the impact of various tunable parameters, optimization methods, and data preprocessing on both the learning speed and the prediction accuracy for the support vector and DNN regressions.

High-Speed Channel Modeling With Machine Learning Methods for Signal Integrity Analysis

Generating eye diagrams by using a circuit simulator can be very computationally intensive, especially in the presence of nonlinearities. It often involves multiple Newton-like iterations at every time step when a SPICE-like circuit simulator handles a nonlinear system in the transient regime. In this paper, we leverage machine learning methods, to be specific, the recurrent neural network (RNN), to generate black-box macromodels and achieve significant reduction of computation time. Through the proposed approach, an RNN model is first trained and then validated on a relatively short sequence generated from a circuit simulator. Once the training completes, the RNN can be used to make predictions on the remaining sequence in order to generate an eye diagram. The training cost can also be amortized when the trained RNN starts making predictions. Besides, the proposed approach requires no complex circuit simulations nor substantial domain knowledge. We use two high-speed link examples to demonstrate that the proposed approach provides adequate accuracy while the computation time can be dramatically reduced. In the high-speed link example with a PAM4 driver, the eye diagram generated by RNN models shows good agreement with that obtained from a commercial circuit simulator. This paper also investigates the impacts of various RNN topologies, training schemes, and tunable parameters on both the accuracy and the generalization capability of an RNN model. It is found out that the long short-term memory (LSTM) network outperforms the vanilla RNN in terms of the accuracy in predicting transient waveforms.

Fast Transient Simulation of High-Speed Channels Using Recurrent Neural Network.

Acoustic noises in the audible range produced by ceramic capacitors has become an increasingly important concern in consumer electronics. The audible noises deteriorate the consumer experience. In order to locate the ceramic capacitors on a printed circuit board (PCB) that cause acoustic noises, we proposed a three-dimensional (3-D) simulation methodology, which is based on the finite element method. The proposed simulation methodology takes into account the product design, the board design, and the detailed component information. The simulation methodology accurately models the board vibrations, which is considered as the primary and the direct cause of acoustic noises. We conduct sound pressure level and laser Doppler vibrometer measurements to benchmark the accuracy and capability of the proposed simulation methodology. On a testing board, the velocity distributions obtained from the proposed simulation agree very well with the measurement results. Through the proposed simulation methodology, we can characterize the board vibrations and extrapolate design guidelines in order to prevent acoustic noise in the early design stage.

Simulation and Characterization of Singing Capacitors in Consumer Electronics

Coupling path characterization is the most critical part of the simulation-based method to investigate desense issues. This paper introduces the workflow to model and validate the coupling path. The workflow is then embodied in a practical case where the GPS system of a phone is severely desensitized by the camera flex. The coupling path between the camera flex and the GPS antenna is successfully characterized and validated by measurement. The obtained coupling path model is further utilized to study factors that potentially affect the noise coupling from the camera flex to the GPS system. Knowledge of these factors facilitates desense design and debugging. Design guidelines are proposed based on the coupling path model.

A Simulation-Based Coupling Path Characterization to Facilitate Desense Design and Debugging

A power distribution network (PDN) is essential in electronic systems to provide reliable power for load devices. With faster load transient current and lower voltage tolerance margin for microprocessors in mobile platforms, it is crucial to optimize the printed circuit board (PCB) design to satisfy the strict target impedance. Conventional modeling methods become impractical in mobile platforms due to the characteristics of high-density interconnect PCB and limited layout space. To overcome these issues, a pattern-based analytical method for the PDN impedance calculation is presented in this article. Based on the localized patterns formulated by the relative relationships between the adjacent vias, parasitic elements are analytically determined for different regions of the entire PCB structure. With the assistance of this method, a practical modeling methodology is developed to construct an equivalent circuit with one-to-one correspondence to the PCB's physical geometry. As a result, the PDN design can be efficiently optimized, especially in the predesign stage, to accelerate the development process. Finally, the proposed method is validated by measurements and full-wave simulations using a real mobile phone PCB in production.

Ken Wu

Papers

High-Speed Channel Modeling With Machine Learning Methods for Signal Integrity Analysis

Fast Transient Simulation of High-Speed Channels Using Recurrent Neural Network.

Simulation and Characterization of Singing Capacitors in Consumer Electronics

A Simulation-Based Coupling Path Characterization to Facilitate Desense Design and Debugging

A Pattern-Based Analytical Method for Impedance Calculation of the Power Distribution Network in Mobile Platforms