Visible Light Positioning (VLP) is a promising indoor localization technology for providing highly accurate positioning. In this work, a VLP implementation is employed to estimate the position of a vehicle in a room using the Received Signal Strength (RSS) and fixed LED-based light transmitters. Classical VLP approaches use lateration or angulation based on a wireless propagation model to obtain location estimations. However, previous work has shown that machine learning models such as Gaussian processes (GP) achieve better performance and are more robust in general, particularly in presence of non-ideal environmental conditions. As a downside, Machine Learning (ML) models require a large collection of RSS samples, which can be time-consuming to acquire. In this work, a sampling scheme based on active learning (AL) is proposed to automate the vehicle motion and to accelerate the data collection. The scheme is tested on experimental data from a RSS-based VLP setup and compared with different settings to a simple random sampling.

In microwave design, Bayesian optimization (BO) techniques have been widely applied to the optimization of the frequency response of components and devices. The common approach in BO is to model and maximize an objective function over the design parameters, in order to find the optimal spectral response. Such an approach avoids the direct modeling of spectral responses, which is a challenging task for the typical data-efficient surrogate models used in BO. Simple objective functions may lead to a suboptimal solutions, while complicated objectives require more powerful and less data-efficient surrogate models. To resolve this issue, this article proposes to adopt a deep Gaussian process (DGP) to directly model all relevant $S$ coefficients over the frequency and the design parameter ranges of interest. Subsequently, an objective probability distribution is retrieved from the DGP model and maximized using a BO scheme. The proposed approach is tested on two suitable microwave examples and compared to the standard BO approach. Results show increased accuracy in identifying the optimal frequency response for the given design parameters and the desired objective, while maintaining high data efficiency.

This work presents a machine learning technique to model wide-band scattering parameters (S-parameters) of interconnects in the frequency domain using a new Gaussian processes (GP) model. Standard GPs with a general-purpose kernel typically assume high smoothness and therefore are not suitable to model S-parameters that are highly dynamic and oscillating due to propagation delays. The new delayed Gaussian process ( $\tau$ GP) model employs a physics-informed kernel consisting of periodic components, whose fundamental frequencies are interpreted as tunable propagation delays. Then, the model hyperparameters are tuned using a combination of maximum marginal likelihood estimation (MMLE) and delay estimation using Gabor transform. The delay estimation allows one to automatically identify the optimal fundamental frequencies for the kernel, thus increasing the numerical stability of the hyperparameters tuning process. The resulting delayed Gaussian process model accurately predicts the S-parameter values at desired frequency points in the training interval. Two application examples demonstrate the increased accuracy of the new technique, compared to standard Gaussian processes, vector fitting (VF), and delayed vector fitting (DVF) rational models.

Engineering design is a complex process to find a suitable trade-off among different, and sometimes conflicting, design specifications. In reality, these requirements can be often considered as constraints of the design problem, that can be defined in terms of performance measures or geometrical characteristics of the device under study. In this paper, a new design space exploration methodology is presented for discovering feasible regions in the design space, where the term feasible region indicates the set of all design configurations satisfying all constraints of the design problem. The proposed method is based on Gaussian process metamodels to estimate the feasible region and leverages a information-based adaptive sampling technique to sequentially refine the prediction accuracy, which is applicable for multiple constraints problems. To efficiently stop the adaptive sampling process, a novel framework to estimate the metamodel’s prediction accuracy is proposed. The efficiency, accuracy and robustness of the proposed approach are compared with state-of-art techniques on suitable benchmark problems and practical engineering examples.

In this article, a new motion-aware sampling strategy (MASS) is presented to speed up the measurement of radiation patterns around cylindrical surfaces. Differently from preexisting sampling techniques, the MASS directly chooses positions that reduce the overall travel time of the field antenna, rather than minimizing the total number of samples. The proposed strategy employs a Gaussian process model that is adapted to the field over a cylindrical surface. Moreover, a new acquisition function for Bayesian active learning is developed in order to efficiently search the peaks of the measured field and predict their values. Next, the proposed strategy is tested on the experimental data from a radiation pattern of a comb generator. Finally, the results are compared to standard grid sampling and Bayesian optimization strategies.

In microwave device and circuit design, many simulations are often needed to find a set of designs that satisfy one or multiple specifications chosen by the designer upfront: the feasible region. A novel Bayesian active learning framework is presented to accurately identify the feasible region with a low number of simulations. The technique leverages on a stochastic model to obtain an efficient and automated procedure. A suitable application example validates the proposed technique and shows its effectiveness to rapidly obtain many suitable designs.