ABSTRACT. The session will discuss how the MATLAB product family supports interactive teaching by showing examples of how MATLAB can interact with different software & hardware products.

ABSTRACT. As Machine Learning as a Service (MLaaS)-based deep learning (DL) solutions are increasingly used in the healthcare industry, concerns regarding personal data privacy have been raised. Consequently, the need for privacy-preserving methods enabling DL model development has grown significantly. Over the past few years, our research has addressed these concerns through various strategies, including homomorphic encryption, encoding, and image obfuscation. In our earlier work, we proposed a variation of a noise-free matrix-based homomorphic encryption scheme (MORE) to ensure privacy in computations within deep learning models. This approach was subsequently employed to create a privacy-oriented cloud-based platform for deploying machine learning algorithms tailored to wearable sensor data. For enhanced security, we integrated a homomorphic encryption scheme that uses modulo operations over integers, an encoding technique that enables computations on rational numbers, and a numerical optimization strategy that facilitates training within a fixed number of operations. To reduce the computational overhead in image-based applications, we proposed several image obfuscation approaches based on non-bijective functions, variational autoencoders, and matrix decomposition. Security assessments were conducted through analytical means and by simulating different attack scenarios. We evaluated the effectiveness of the proposed approaches by implementing them in diverse medical applications, achieving promising results.

ABSTRACT. The work presents some results of signal analysis, periodic and aperiodic behavior, with artificial intelligence techniques. Two paradigms are considered, based on artificial neural networks (ANN) and – the second one – based on particle swarm optimization (PSO). The first approach promotes linear neural networks architecture, ADALINE for single signals and MADALINE for multiple signals case. The second approach promotes the swarm intelligence, used here for the minimization of objective function based on mean squared error (MSE). For periodic signals, the trigonometric Fourier series is considered. For aperiodic signals, the discrete Fourier transform is promoted. Both techniques could be applied for a more general problem of model parameter estimation. The computer-based experiments are made with various signal-to-noise ratios, to study the effects on the converge rate on the estimation process. At least for the considered signals, the ANN based solution provides better results. The framework could be extended to the analysis problem based on other base functions, e.g., wavelets.

ABSTRACT. The aim of the work is to make an evaluation of two versions of Particle Swarm Optimization (PSO), in the framework of signal analysis and model parameters estimations. The versions of PSO are with constraint coefficients and inertia (PSO-COIN) and the version based on quantum models (QPSO). Both versions are members of the evolutionary computation domain and have bioinspired roots. Two case studies are considered. The first case study is represented by two test functions, chosen from a benchmark suite used for the evaluation of the optimization algorithms. The second case is the parameter estimation problem, in the context of the signal analysis framework, and under various signal-to-noise ratios to study the effects on the converge rates of the estimation processes. The computer-based experiments show comparable performances for PSO algorithms, for short time simulation, e.g., 100 iterations. After that, the version of PSI-COIN gives better results.

ABSTRACT. The objective of the paper is to present a way to develop and train a neural network in Matlab, in order to run it on a customized evaluation board. The essence of the development of this route is, on the one hand, the fact that the customized Pumpkin board has associated software dedicated to running artificial neural networks on the Graphics Processing Unit - GPU, which gives a high speed of simulating neural networks thanks to the parallel processing feature and on the other hand, the Matlab environment has important resources for the development, characterization and optimization of artificial neural networks - ANN. The considered artificial neural network is of the "deep network" category, with fifteen layers and the data structure is specific to this category, being different from the "shallow network" category. The data presented for training and running the network is of image type with eight-bit greyscale representation and consists of digits from 0 to 9. The classification results made by the network are compared for running in Matlab as well as on the considered Pumpkin board.

ABSTRACT. This paper introduces V-CNN, a versatile structure inspired from previously introduced light models such as NL-CNN. What makes the V-CNN structure efficient is the process of optimizing its hyper-parameters. Consequently, several aspects in proper design for building of efficient (validation accuracy near state of the art while keeping the complexity of the structure) V-CNN structures, are considered and detailed in this paper. While V-CNN includes as particular cases several previously defined models such as NL-CNN and XNL-CNN, it can be better tailored for efficient deployment given a specific dataset. A V-CNN model with 1.5 million parameters obtained 91.55% validation accuracy on CIFAR-10 dataset, surpassing the previous result of 90.60% using the NL-CNN. The V-CNN model offers the following advantages: i) using optimization hints described herein it allows maximal efficiency (good accuracy at low complexity) for a wide variety of datasets; ii) has a low number of layer primitives, thus making easier their specific design for deployment on various TinyML or EdgeAI platforms, including FPGAs.

ABSTRACT. More than homogenous or heterogenous unmanned vehicles can work in a coordinated manner and perform large-scale swarm tasks (firefighting, search and rescue, mapping, and military operations, etc.) efficiently in a shorter time by sharing their workload. The collision of these vehicles is one of the most significant challenges they face during their operations. The crash of the vehicles causes the vehicles to be out of duty and the mission's failure. In this study, Quadrotor-type UAVs used as homogenous agents can go to any target point by receiving location, speed, and compass information with GPS and IMU sensors. In the application, the positions of the agents were kept and regularly updated in a paired list, resembling the traversal process found in the Optimized Bubble Sort Algorithm. The experiments were conducted on swarm agents with the graph-based collision avoidance algorithm and fully autonomous swarm capabilities. The results show that the proposed method can effectively avoid collisions and navigate swarm in the simulation environments.

ABSTRACT. This paper introduces XNL-CNN, an extended version of the previously introduced, compact NL-CNN model. While the NL-CNN was designed with typical Tiny-ML constraints in order to make possible integration of various AI image recognition solutions into low power, low complexity computing platforms, the same stands for its extended version discussed herein. In addition, we consider additional layers to cope with larger image sizes, as found in most actual relevant datasets (e.g. the Cars196 or Flowers104). Moreover, since working with large image sizes requires intensive computational power, TPU accelerators readily available on Kaggle platform were used for training and evaluation of our model. In comparison to other, widely known resources-constrained architectures, such as EfficientNet or MobileNet our model offers the following advantages: i) a larger set of hyper-parameters allowing the user to do better designs (maximal accuracy for minimal resources), adapted for a wide variety of datasets; ii) has a low number of layer primitives, thus making easier their specific design for deployment on various TinyML or EdgeAI platforms, including FPGAS.

ABSTRACT. The paper presents the changes in the thermal regime of rewound induction motors to obtain other functional characteristics than those established, initially, through the design process. Due to distribution of temperatures within the structure of the induction motor - using Motor-CAD programming software - it is highlighted that, under the conditions of compliance with the claimed requirements for maintaining both the strength of the magnetic field (also the forces developed by the induction motor) as well as the current density, when rewinding to obtain other functional characteristics on a given physical structure/geometry of the squirrel cage induction motor, the thermal regime is different from the reference version, that can radically influence the duration of its life.)

ABSTRACT. Abstract – In this paper, electromagnetic transient processes in a three-phase cage induction motor due to sudden disconnection and subsequent fast reconnection of its electric power supply are analyzed using linear complex differential equations of the representative-space-phasor electromagnetic model of the motor. With this in view, it is assumed that the disconnection of the motor from the electric power supply (before its reconnection to the same power supply by rapid automatic reclosing) has a duration of just a few cyclic time-periods of the three-phase time-harmonic supply voltage system, so that, the motor speed remains unchanged over this very short duration. The electromagnetic transient analysis is corroborated by meaningful case-study numerical results.

ABSTRACT. Given the massive movement towards hybrid power sources and electrification in the marine industry, power conversion systems are increasingly needed to support the flexibility and efficiency required by shipboard electrical grid. Electromagnetic compatibility has been considered as a matter of concern for special ships and for navy designs. The increase in the volume of power electronics systems relative to the installed power onboard the ship brings EMC compliance as a key factor for the good operation of the electrical installation. This paper addresses some EMC aspects that need to be considered in the design and installation of electrical systems incorporating power electronics equipment. Examples are used to highlight some of the solutions used to limit the electromagnetic interference that may occur during the operation of this equipment.

ABSTRACT. Most of the connection and control schemes for connecting inverters to the network propose for MPPT tracking the connection of a Boost converter connected to the inverter in the power circuit. Another way of coupling is the single stage topology in which the MPPT tracking is also of the Boost converter type to obtain the reference voltage. Most control strategies use the transformation of three-phase quantities in Synchronous Reference Frame Control and power control. This paper deals with strategic control of the DC-link voltage in Stationary Reference Frame Control (SRFC) with the single stage topology in which the MPPT tracking is the Buck converter type to obtain the reference voltage. The control strategy applied to the grid-side converter consists mainly of two cascaded loops. Firstly, there is a fast internal current loop responsible for power quality, which regulates the grid current, and an external voltage loop, which controls the dc-link voltage. The dc-link voltage controller is designed for balancing the power flow in the system. The design of this controller aims for system stability having slow dynamics having where the proportional gain is realized by damping controller(negative gain) ,which is a renewed controller to control the grid voltage and current based on negative imaginary approach. Based on this control strategy, the simulations regarding the operation of the inverter under Standard Test Conditions (STC) of irradiance of 1000 W/m2, spectrum AM 1.5 and cell temperature of 25°C, indicate a very good response of the inverter with a high efficiency. The tests in temperature limit conditions (-10oC and +70oC) performed under STC and Nominal Module Operating Temperature (NMOT), irradiance of 800 W/m2, satisfy the requirements for good operation of the inverter.

ABSTRACT. With the new feature MATLAB Compiler™ available starting R2020b Matlab, an application developed in Matlab can be packaged as docker standalone application and deployed using Docker® container service. The target system running the standalone application must be a Linux distribution and requires a MATLAB® Runtime installation and an active docker service. The proposed standalone application is a PV simulator GUI programmatically implemented in Matlab based on two photovoltaic (PV) panels’ data sheets. The obtained PV simulator is launched in both MATLAB® environment, where initial was designed and externally in a Linux environment as a standalone application independent of the design program. No differences between the reported results, this proving the possibility of large-scale usability of what was designed.

ABSTRACT. The increasing complexity of today's electronic systems requires efficient information exchange between their components. This is often achieved using shared memory locations, which in SoC devices, microcontrollers and other hardware components are called registers. This paper demonstrates how software mechanisms called semaphores can efficiently accomplish this information exchange. Long distance communication between electronic systems is also addressed, and a complex solution for transferring information over the Internet using state-of-the-art technologies is presented. In addition, since many critical applications require real-time functionality, this study explains both theoretically and practically the advantages of using a real-time operating system. For a better understanding of the concepts presented, the information in the article is validated by creating from scratch a physical device having a pet feeder functionality, designing its circuitry, software architecture and physical structure.

ABSTRACT. In the context of increasing car traffic, with significant particularities in EU countries in terms of road infrastructure, advanced driver assistance systems have been developed, based on technical developments in sensors, electronics and information and communication technology. Electronic computing units (ECUs) and microcontroller units process the information received from sensors based on algorithms implemented to synthesize the vehicle's environment in real time. Advanced Driver Assistance Systems (ADAS) are integrated, artificial intelligence-based technologies aimed at increasing safety and security for traffic participants. ADAS features aim to reduce/avoid accidents/collisions and reduce the severity of trauma to the personnel involved by providing information to the driver about potentially dangerous situations caused by human error and to improve the quality of driving. Human error is one of the most important factors influencing the severity and number of accidents. The technique of automatic transmission and processing of information allows monitoring the driver (for identification and classification of driving style) in relation to a series of criteria that contribute to increasing traffic safety, improving comfort, optimizing fuel consumption, reducing pollution, etc. by generating algorithmically generated warnings. This paper presents the architecture of an intelligent automated vehicle operator traffic monitoring system (SiaMOTO) which, through the intelligent automated vehicle operator traffic monitoring devices (DiaMOTO) it incorporates, actively tracks the behavior of drivers of a target fleet of vehicles and analyses their traffic behavior characteristics. The SiaMOTO/DiaMOTO infrastructure is a special system set up to run AI/DLM/BigDATA algorithms through which systematic errors of monitored drivers will be detected and will allow, after complex analyses, to generate preventive warnings. SiaMOTO is an information environment with a preventive role, based on the learning of systematic characteristics of the behavior of the drivers it assists in traffic.

ABSTRACT. In this paper the results of research in optimizing a waveguide antenna, by applying optimization techniques based on genetic algorithms, are presented. The optimization process consists in identifying the radiating elements that have to be eliminated from a fully populated configuration (elements that cancel receive the excitation value equal to 0, and those that remain active receive the excitation value equal to 1).

ABSTRACT. This paper aims to explore the design of an FPGA based 8 channel data-acquisition device. The hardware and hardware-design-language described in this paper, are purpose-built for a select set of components. These components are selected to guarantee a minimum data acquisition rate of \SI[per-mode=symbol]{30}{\mega\sample\per\second} per channel. This paper will detail a general system overview and part selection, followed by the tested part of this system.

ABSTRACT. In wireless communications the diversity techniques are used with success to mitigate de perturbations from the transmitted channel. These interferences can be generated by different sources and can be noises or multipath fading effect. The noise can be Gaussian or impulsive with various mathematical models. The impulsive noise degrades sever the signal quality, affecting groups of successive bits in data streaming. Using interleaver this effect can be decreased by rearrange the bit sequence before encoding. In this paper an Alamouti code with interleaver system is proposed to mitigate the impulsive noise with SαS distribution. Alamouti code is a spatial diversity technique, and it has good performances on Additive White Gaussian Noise channel. Data are transmitted on multiple antennas at the emitter and at the receiver. We considered 2x2 system with Binary-Phase-Shift-Keying modulation, Rayleigh fading and two types of interleavers: random and Takeshita-Costello. Both of them has smaller error rate, but the best is the last one. The results are compared with gaussian noise. The model of impulsive noise is described and analyzed for different values of its parameters.

ABSTRACT. This paper combines the fields of radiofrequency (RF) electronics and automated tuning using a microcontroller in order to optimize the process by continuous adjustment of the parameters. We propose an experimental setup composed of several RF stages, parts of a radio communication chain, a control system implemented on ATMEGA 328p microcontroller that uses 3 digital ports to generate PWM control signals and an analog input port to read the value used for the optimization process. The idea finds its usefulness in testing new antenna technologies and can be used for automated testing of the emission or reception efficiency. The test equipment consists of 2 antennas that communicate with each other using radio waves and can be considered as coupled by a waveguide that isn’t uniform. In this case the improper tuning of the wave guide with the other components of the communication chain generates power losses due to reflections and damage to the equipment. Because tuning depends on the reflection coefficient that is influenced by the component’s impedances, any transmission line has a certain propagation structure at a given frequency. In this paper we propose the inspection of a transmitter/receiver chain in the context of controlled variation by a microcontroller of the frequency and power of transmission under certain tolerable limits. Graphs with the obtained results are presented.