Volumen 15 (2019): Heft 2 (December 2019)

In the paper, the problem of building semantic image segmentation networks in a more efficient way is considered. Building a network capable of successfully segmenting real-world images does not require a real semantic image segmentation task. At this stage, called prototyping, a toy dataset can be used. Such a dataset can be artificial and thus may not need augmentation for training. Besides, its entries are images of much smaller size, which allows training and testing the network a way faster. Objects to be segmented are one or few convex polygons in one image. Thus, a toy dataset generator is created whose complexity is regulated by the number of edges in a polygon, the maximal number of polygons in one image, the set of scale factors, and the set of probabilities determining how many polygons in a current image are generated. The dataset capacity and image size are concurrently adjustable, although they are much less influential.

The paper aims at researching and developing an adaptive control system algorithm and its implementation and integration in the control system of the existing unmanned aerial vehicle (UAV). The authors describe the mathematical model of UAV and target function for energy consumption minimisation and possible searching algorithms for UAV optimal control from an energy efficiency perspective. There are two main goals: to minimise energy consumption and to develop and investigate an adaptive control algorithm for UAV traction drive in order to increase energy efficiency.

The optimal control algorithm is based on two target function values, when comparing and generating corresponding control signals. The main advantage of the proposed algorithm is its unification and usability in any electrical UAV with a different number of traction drives, different or variable mass and other configuration differences without any initial manual setup. Any electric UAV is able to move with maximal energy efficiency using the proposed algorithm.

Unmanned aerial vehicles (UAVs), especially drones, have advantages of having applications in different areas, including agriculture, transportation, such as land use surveys and traffic surveillance, and weather research. Many network protocols are architected for the communication between multiple drones. The present study proposes drone-following models for managing drones in the transportation management system in smart cities. These models are based on the initial idea that drones flight towards a leading drone in the traffic flow. Such models are described by the relative distance and velocity functions. Two types of drone-following models are presented in the study. The first model is a safe distance model (SD model), in which a safe distance between a drone and its ahead is maintained. By applying the stochastic diffusion process, an improved model, called Markov model, is deduced. These drone-following models are simulated in a 2D environment using numerical simulation techniques. With the simulation results, it could be noted that: i) there is no accident and no unrealistic deceleration; ii) the velocity of the followed drone is changed according to the speed of the drone ahead; iii) the followed drones keep a safe distance to drone ahead even the velocities are changed; iv) the performance of the Markov model is better than that of the SD model.

The article discusses vulnerability of wireless sensors networks to weather-based disruptions considering the opinions of different experts published in a range of scientific materials. The introduction provides a brief overview of wireless signals in real world conditions focusing on how weather affects signals (rain, fog and clouds, snow, hail, lightning, wind, bodies of water, trees and physical obstruction). Information about the effects of weather on wireless sensor networks using Free Space Optical / Radio Frequency (FSO/RF) communication is then provided. Finally, the impact of weather conditions on MANET routing protocols is considered theoretically, and experimental simulations are performed by comparing the sustainability of different protocols to different weather conditions. After analysis of experiment results, ideas on how to decrease vulnerability of wireless networks to weather-based disruptions are discussed.

In both industrial and scientific frameworks, free and open source software codes create novel and interesting opportunities in computational electromagnetics. One of the possible applications, which usually requires a large set of numerical tests, is related to antenna design. Despite the well-known advantages offered by open source software, there are several critical points that restrict its practical application. First, the knowledge of the open source programs is often limited. Second, by using open source packages it is sometimes not easy to obtain results with a high level of confidence, and to integrate open source modules in the production workflow. In the paper, a discussion about open source programs for antenna design is carried out. Furthermore, some preliminary numerical tests are presented and discussed, also in comparison with those obtained by means of commercial software. Results are related to the simulation of various typologies of antennas in order to assess the capabilities of open source software in different configurations. The presented comparisons show that, despite the abovementioned limitations, the examined open source packages have similar performance with respect to their commercial counterparts.

The paper addresses the impact of peak-to-average power ratio (PAPR) and spectrum of the waveform, as well as load resistance on the performance of low-power harvesting device in a real-life wireless power transfer (WPT) scenario. In the current study, a combination of the classic voltage doubler circuit for RFDC conversion and premanufactured device for DC-DC conversion is used. For the investigation of conversion efficiency and harvesting device performance, three types of waveforms are used: single tone, multicarrier signals with low PAPR and multicarrier signal with high PAPR. In order to generate high-PAPR signal, subcarriers with the same amplitude and phase are summed, whereas for generation of low PAPR signal the phases of the subcarriers are chosen pseudo-randomly. Over-the-air transmission in 865 MHz ISM band is made using directional antennas and all multicarrier waveforms have equal 5 MHz bandwidth. To evaluate the performance of harvesting device and conversion efficiency, the average voltages at the input and output of the RF-DC converter as well as at the output of the DC-DC converter with corresponding input and load impedance are measured. The experiments have shown that the employed multicarrier signals can greatly improve the performance of harvesting device during WPT under certain conditions, which are discussed in the paper.

Power transformer is one of the most important pieces of equipment in the grid to reliably and efficiently transmit power to the consumers. Asset management and protection are the best concepts for prolongation of transformer lifespan as well as for the increase of grid reliability. This article presents electrical and non-electrical parameter-based power transformer monitoring and protection. Various data such as core flux, age of asset, heat generation, current harmonics and temperature are monitored in real time and processed accordingly to enhance the working capability of the transformer. The proposed scheme is successfully tested on 15 kVA laboratory transformer using Arm CORTEX-M4 processor. A Fitness Function (Ff) is estimated from the collected data to examine the working condition of the transformer. Moreover, voltage, current and power-based inrush detection as well as Adaptive Power Differential Protection (APDP) are applied to protect the transformer against fault. The hardware implementation and result validation prove the effectiveness of the proposed scheme in enhancing reliability of the distribution grid.

The work reports on the design and performance of a low-noise low-cost CMOS transimpedance amplifier (TIA). The proposed circuit shall be employed in optical time-domain reflectometers and is implemented using an affordable 0.18 µm 1.8 V CMOS process. The approach preserves the benefits of a classical feedback structure while addressing the noise problem of conventional feed-forward and resistive feedback architectures via the usage of noise-efficient capacitive feedback. Circuit-level modifications are proposed to mitigate the voltage headroom and DC current issues. The suggested design achieves a total gain of 82 dBΩ (79 dBΩ after the output buffer) within the bandwidth of 1.2 GHz while operating with a total input capacitance of 0.7 pF. The simulated average input-referred noise current density is below 1.8 pA/sqrt(Hz) with the power consumption of the complete amplifier including the output buffer being 21 mW.