Applications in Electronics Pervading Industry, Environment and Society

Table of Contents

1. Machine Learning

   1. Frontmatter

   2. Affordance Segmentation Using RGB-D Sensors for Application in Portable Embedded Systems

      Edoardo Ragusa, Matteo Pastorino Ghezzi, Rodolfo Zunino, Paolo Gastaldo

      Abstract

      Depth sensors play a major role in the control pipeline of semi-autonomous prostheses. The feed from these sensors can complement, or even substitute, the information retrieved by RGB cameras. The paper explores the application of depth sensors to affordance detection, to recognize the graspable part of an object in foreground images. The experiments confirm that the depth (“D”) components can boost the accuracy in predicting the affordable parts of an object, when they are matched with state-of-the-art computer vision architectures. A prototype of a portable inference system with real-time performances has been assembled using a Jetson Tx2 and a Realsense D435i camera, confirming the proposal’s suitability for semi-autonomous prostheses.

   3. ML-Based Classifier for Precision Agriculture on Embedded Systems

      Romina Soledad Molina, Valentina Carrer, Maynor Ballina, Maria Liz Crespo, Luciana Bollati, Daniel Sequeiro, Stefano Marsi, Giovanni Ramponi

      Abstract

      In precision agriculture, effective pest control helps to reduce yield loss and pesticide application. In this research, the pest to be detected and controlled is the moth lobesia botrana, which mainly attacks the vineyard. We present an automatic pest classifier based on machine learning, considering resource-constrained devices in IoT systems. Transfer learning and an ensemble of compression techniques are used to reduce the size of the classifier with a good trade-off between efficiency, effectiveness, and resource utilization. This procedure allows the achievement of a fully on-chip deployment in two technologies: esp32 and SoC-based FPGA Xilinx PYNQ-Z1 and KRIA.

   4. Deep Reinforcement Learning for Automated Car Parking

      Luca Lazzaroni, Francesco Bellotti, Alessio Capello, Marianna Cossu, Alessandro De Gloria, Riccardo Berta

      Abstract

      This article explores the development of a Deep Reinforcement Learning (DRL) -based agent able to perform both path planning and trajectory execution, processing sensor perception information and directly controlling the steering wheel and the acceleration, like a normal driver. As a preliminary investigation, we limit our research to low-speed manoeuvers, in a challenging narrow drivable area. The vehicle’s agent completely relies on the real-time information from the sensors, thus avoiding the need of a map. We show the validity of the proposed system in a simulated car parking test, in which the agent has been able to achieve high target reach rates, with a limited number of manoeuvers (gear inversion rate), outperforming the well-established Hybrid A-Star path planning algorithm in both the metrics. Further research is needed for improving the generalization ability of the agent and its application in more dynamic driving environments.

   5. Machine Learning Techniques for Anomaly-Based Detection System on CSE-CIC-IDS2018 Dataset

      Abdussalam Elhanashi, Kaouther Gasmi, Andrea Begni, Pierpaolo Dini, Qinghe Zheng, Sergio Saponara

      Abstract

      Anomaly-based detection is a novel form of an intrusion detection system, which has become the focus of many researchers for cybersecurity systems. Data manages most business decisions. With more access to data, it is necessary to interrupt and analyze them correctly. When it comes to security, the first step is to determine the outliers as a security threat. Machine learning and deep learning techniques have proven to recognize anomalous attack patterns that deviate from normal network behavior. Machine learning can be utilized to learn the characteristic of data and help to improve the speed of detection. In this research, we present our approach to implementing an algorithm for the anomaly detection framework in complex and unbalanced data. The proposed method has been applied to a CSE-CIC-IDS2018 dataset. It is the most recent dataset that is publicly available, an extensive dataset that includes a wide range of attack types. This data has been pre-processed and cleaned to find helpful information for classification by the proposed models. We performed a correlation methodology to filter irrelevant anomalies and grouped the correlated anomalies into a single feature to minimize detection time. A stacked autoencoder has been used to reduce the dimensionality of the dataset. We exploited different machine learning algorithms such as (Random Forest, GaussianNB, and multilayer perceptron) to classify the streamed data. Our experimental results outperformed the superiority of the proposed approach to identify anomalous components and manage threat detection in cybersecurity applications.

   6. AI-Based Sound Event Detection on IoT Nodes: Requirements Evaluation

      D. Errico, M. Re, V. Colombo, G. C. Cardarilli, M. Martina, M. Ruo Roch

      Abstract

      Sensors capable to detect specific audio events can be deployed inside smart cities, to improve citizens safety. A sensor cloud would allow to georeference relevant incidents, like screams, car crashes, or gunshots. An IoT based approach requires the development and deployment of smart nodes combining minimal power consumption and reasonable preprocessing capabilities, to minimize both power supply requirements and the amount of transmitted data. In this work, a possible system architecture is presented, and a detailed analysis of IA approaches to sound event detection is carried-out. Optimizations for IoT nodes deployments are then applied, and a performance comparison to current algorithms is presented.

   7. Contextual Bandits Algorithms for Reconfigurable Hardware Accelerators

      Marco Angioli, Marcello Barbirotta, Abdallah Cheikh, Antonio Mastrandrea, Francesco Menichelli, Saeid Jamili, Mauro Olivieri

      Abstract

      Reconfigurable processing cores for IoT and edge computing applications are emerging topics to calibrate costs, energy consumption and area occupation with performance and reliability on Commercial Off the Shelf (COTS) devices. This work analyzes how to take advantage of Machine Learning to potentially automate the reconfiguration process of a hardware accelerator inside the Klessydra Vector Coprocessor Unit (VCU), choosing the best configuration according to the workload. The problem is modeled with a contextual bandits approach using the Linear UCB algorithms and validated with offline Python simulations.

2. Transport, Energy, Security, Health

   1. Frontmatter

   2. Low-Cost Lithium-Ion Battery Characterization Setup Based on Auxiliary Batteries

      Niccolò Nicodemo, Roberto Di Rienzo, Alessandro Verani, Federico Baronti, Roberto Roncella, Roberto Saletti

      Abstract

      Characterization tests are key to improve the Lithium-Ion batteries performance. Unfortunately, they require expensive equipment that small companies and laboratories can hardly afford. In this paper a novel low-cost battery characterization setup is proposed. It uses an auxiliary battery to strongly reduce the required maximum instrument power and cost. A simulation framework of the proposed setup is developed in Matlab/Simulink environment and is applied to the characterization of a 48 V mild hybrid battery. This case study is used to both explore setup advantages and guide through its sizing. The obtained results show a strong reduction of the power required to the test equipment and a reduction up to about 70% of the energy drawn from the grid.

   3. Model-Based Vital Control Architecture for Highly Automated Train Operations

      Giovanni Mezzina, Cataldo L. Saragaglia, Mario Barbareschi, Diana Serra, Salvatore De Simone, Alberto Moriconi, Daniela De Venuto

      Abstract

      The railway panorama is experiencing notable development due to the introduction of always new infrastructures with high grade of automation, realizing the so-called ATO over ETCS (AoE) framework on mainlines. Despite AoE achieved optimal results for supervised operations, unattended train operations management is still in an embryonal stage. In this context, this paper proposes a model-based architecture, Vital Control Module (VCM), that improves the safety of unattended convoys operating on infrastructures with highly automated AoE. The model, developed in Matlab/Simulink, supervises the framework operativity, estimates train speed and communicates with the trackside-connected operator. Exploiting these functions, VCM can substitute ETCS when faulty or unpowered, and detect hazardous situations intervening in a negligible time. To test the model, a custom and interactive emulator has been realized. Experimental results showed that the VCM can detect hazards and mitigate them in 451 ms–533 ms, halving the time required by the related standard constraints.

   4. Exposure of the Human Head to 5G Electromagnetic Radiations: Modeling and Analysis

      Sara Alameddine, Dina Al-Houmsy, Ali Mohsen, Houssein Hajj Hassan, Ali Ibrahim, Mohamad Hajj-Hassan

      Abstract

      5G, the 5th generation cellular network provides new utilities to different kind of business and industry sectors such as healthcare, electronics, and communications. In addition, it provides numerous features that distinguish it from previous generations such as high speed of internet reaching 10 Gbps, increase in the amount of data transmitted “wide bandwidth”, and advanced antennas. Although 5G has promising future, it still has a blurry side that disquiet most of the scientists, since there are no final studies ensuring that the frequency of 5G has no effects on human health especially on the brain, additionally that the people usage of the cellphone will reach its peak with 72.6% by 2025. In this paper, numerical simulations of different ranges of frequencies are conducted on a new designed model according to different international standards representing the biological, physical, and chemical characteristics of the sub-layers of the human head, to investigate the effect of 5G radiations on the head. Here, we investigate the effects of Specific Absorption Rate (SAR) averaging mass, the heat that results from other sources, and the propagation of the wave on the correlation with temperature elevation under the effect of 5G exposure over time. The obtained results are employed to check the safety of the simulated scenario.

   5. A Short-Range Free-Space Optical Communication System for Space-Assembled Microsatellites

      Demetrio Iero, R. Carotenuto, M. Merenda, F. G. Della Corte

      Abstract

      A free-space optical transmission system for applications that require a proper data communication channel between docked microsatellite modules is presented. The structure and the design of the optical communication interfaces are described. The system acts as a bridge between the CAN (Controlled Area Network) busses on the microsatellites, allowing communication between circuits in different separated modules. This avoids a physical connection between adjacent modules and electromagnetic disturbances that radio-frequency communication systems can generate. The system uses low-cost Commercial-Off-The-Shelf (COTS) components that can reduce significantly the operative costs. A prototype has been built in a format compatible with CubeSat satellites and successfully characterized and tested.

   6. A Low-Area, Low-Power, Wide Tuning Range Digitally Controlled Oscillator for Power Management Systems in 28 nm CMOS Technology

      M. Mestice, G. Biondi, G. Ciarpi, D. Rossi, S. Saponara

      Abstract

      Nowadays, in the world of high-performance computing, saving energy when great computing power is not needed is a must-to-have feature. This usually involves the implementation of Power Management Systems (PMS) to apply power saving polices such as frequency scaling. In particular, for this feature, the actuators of PMS are usually implemented with Phase- or Frequency-Locked Loops, which should occupy a small area and exhibit a low-power consumption. Additionally, they should be able to generate a wide range of frequencies in the order of a few GHz with a fine granularity of a few hundreds of MHz. Since the core of such loops is a tunable oscillator, in this work we present a pseudo-differential Ring Digitally Controlled Oscillator (DCO) implemented with a standard 28 nm CMOS technology to be used in PMS. The proposed DCO features a well-balanced behavior between the noise performance and a wide tuning range, a low-area, and a low-power consumption.

   7. Sorting of Live/dead Escherichia Coli by Means of Dielectrophoresis for Rapid Antimicrobial Susceptibility Testing

      A. di Toma, G. Brunetti, N. Sasanelli, M. N. Armenise, C. Ciminelli

      Abstract

      According to the World Health Organization (WHO) forecasts, AntiMicrobial Resistance (AMR) will represent the leading cause of death worldwide in the next decades. To prevent this phenomenon, rapid antimicrobial susceptibility testing is needed to guide the choice of the proper antibiotic. In this context, we propose a chip-scale system, mainly based on a microfluidic channel combined with a pattern of engineered electrodes, to efficiently test an antibiotic on a bacteria sample. The use of dielectrophoretic (DEP) forces enable the sorting of live/dead bacteria, such as Escherichia Coli, with an efficiency larger than 99% for rapid monitoring of the antimicrobial susceptibility at the single-bacterium level.