Accurate prediction of dynamic states in complex physical systems is critical for intelligent decision-support and control. This paper presents a novel spatio-temporal deep learning framework tailored to expert systems operating in mobility and transportation environments. Using a large-scale, multi-source dataset collected over two years from a 100-kilometer highway segment, the proposed model integrates static road geometry features—specifically curvature and slope—into a Long Short-Term Memory (LSTM) architecture. We introduce GeoLSTM, a geometry-aware neural network model that enhances mid-term traffic forecasting accuracy across multiple metrics, including vehicle speed, total intensity and heavy-vehicle (HV) intensity, addressing the common omission of static road geometry by embedding curvature and slope. Our experiments demonstrate that incorporating static spatial features significantly reduces prediction error (up to 26 % for speed and 16 % for heavy-vehicle intensity) and improves model stability, outperforming both traditional baselines and transformer-based alternatives. We further quantify computational costs to assess real-time deployment feasibility. These results establish a robust methodology for integrating contextual spatial data into deep learning-based expert systems, with potential generalization across other domains requiring predictive modelling in structured, data-rich environments.

In interferometric radiometry, the correlations between all pairs of radio-frequency (RF) receivers must be performed in order to obtain the scene visibilities. This represents a cumbersome problem in passive imaging where large signal bandwidths are typically required to achieve fast-acquisition times and improve the radiometric resolution of the image. In this case, the signal distribution and the correlation at intermediate frequencies require very fast signal acquisition and processing subsystems. This paper presents a technique to perform correlations of millimeter-wave signals produced by thermal emission in the optical domain as a solution to the aforementioned problem. The proposed method is based on converting the RF signal to the optical domain by modulating a laser beam with the RF signal using a LiNbO3phase modulator. This conversion allows to perform the signal distribution in the optical domain to obtain the combination of the receiver pairs. The correlation is obtained by measuring the power of the photocurrent produced by photodetecting the combined signal. The results of an experimental validation consisting on the acquisition of passive images using a linear interferometric array are presented to support the feasibility of the method. In addition, performance considerations of the system have been developed and validated by calculating the standard deviation of a visibility measurement.

This paper presents a comparative study of the radiometric sensitivity and spatial resolution of three near-field (NF) passive screener systems: real aperture, 1-D synthetic aperture (SA), and 2-D SA radiometers are compared. The analytical expressions for the radiometric resolution, the number of required antennas, and the number of pixels in the image are derived taking into account the distortion produced by the NF geometry at nonboresight directions where the distortion is dominant. Based on the theoretical results, a performance comparison among the studied systems is carried out to show the advantages and drawbacks when using the radiometers in a close-range screening application. Moreover, the screener performance in a close-range environment is discussed from the results obtained in the aforementioned comparison.

Radio-frequency identification (RFID) technology relies on the modulated scattering technique (MST) as a means to convey the information from the tag to the reader. The interaction mechanism between the reader and the tag is described in a physically meaningful way by the reciprocity theorem which is deeply rooted into Maxwell equations. This approach provides a simple and yet complete formulation that allows to fully describe the interaction between the reader and the tag, even in complex environments. From this formulation, a clear understanding is derived on how the different design tradeoffs affect system performance. Based on this approach, the paper is focused on several issues of practical relevance for antenna tag design such as maximum power transfer, maximum sensitivity, and nonlinearity effects.

This paper proposes low overhead training techniques for a wireless communication system equipped with a Multifunctional Reconfigurable Antenna (MRA) capable of dynamically changing beamwidth and beam directions. A novel microelectromechanical system (MEMS) MRA antenna is presented with radiation patterns (generated using complete electromagnetic full-wave analysis) which are used to quantify the communication link performance gains. In particular, it is shown that using the proposed Exhaustive Training at Reduced Frequency (ETRF) consistently results in a reduction in training overhead. It is also demonstrated that further reduction in training overhead is possible using statistical or MUSIC-based training schemes. Bit Error Rate (BER) and capacity simulations are carried out using an MRA, which can tilt its radiation beam into one of Ndir= 4 or 8 directions with variable beamwidth (≈2π/Ndir). The performance of each training scheme is quantified for OFDM systems operating in frequency selective channels with and without Line of Sight (LoS). We observe 6 dB of gain at BER = 10-4and 6 dB improvement in capacity (at capacity = 6 bits/sec/subcarrier) are achievable for an MRA with Ndir= 8 as compared to omni directional antennas using ETRF scheme in a LoS environment.

In this paper, the capability of ultra-wide-band (UWB) sensor arrays for tomographic radar of electrically large objects is presented. The major concern when imaging is extended to real objects is to achieve a correct reconstruction of the object shape and its electric properties. A general framework based on a UWB bifocusing operator (UWB-BF) with good tomographic imaging capabilities is presented. This general approach provides a comprehensive understanding of the basic tradeoffs with regard to sensing geometry and image quality parameters. Through numerical simulations and measurements applied to canonical as well as to complex objects, basic design criteria are assessed and the potential of UWB tomographic radar imaging is presented.

A compact system for a three dimensional (3D) pointer (mouse) is possible by combining microwave holography with microstrip antenna arrays to enter 3D coordinates of a point in space into computer memory. To reach the practical reduced size system, new solutions are proposed in both the receiving planar array antenna for constructing the point hologram image and the holographic algorithms for reconstructing the microwave hologram.

A comparative study at 2.45 GHz concerning both measurement and reconstruction parameters for planar and cylindrical configurations is presented. For the sake of comparison, a numerical model consisting of two nonconcentric cylinders is considered and reconstructed using both geometries from simulated experimental data. The scattered fields and reconstructed images permit extraction of very useful information about dynamic range, sensitivity, resolution, and quantitative image accuracy for the choice of the configuration in a particular application. Both geometries can measure forward and backward scattered fields. The backscattering measurement improves the image resolution and reconstruction in lossy mediums, but, on the other hand, has several dynamic range difficulties. This tradeoff between forward only and forward-backward field measurement is analyzed. As differential temperature imaging is a weakly scattering problem, Born approximation algorithms can be used. The simplicity of Born reconstruction algorithms and the use of FFT make them very attractive for real-time biomedical imaging systems.