Search

Resource allocation in vehicular communication often encounters stringent constraints that are hard to satisfy due to high mobility and a complicated communication environment, making the optimization problem infeasible. However, even in such infeasible scenarios, the resource allocation strategy should be able to provide viable solutions within a short computation time. To address this challenge, we explore the potential of a deep learning (DL) framework that can provide reasonable resource allocation solutions even when certain constraints cannot be satisfied. In particular, we focus on resource allocation to maximize overall energy efficiency while ensuring minimum spectral efficiency, where resource allocation constraints may not always be satisfied, unlike traditional works that consider only the feasible scenarios. We propose a DL framework that uses deep neural network (DNN) models to approximate resource allocation. In addition, an unsupervised learning-based training methodology is developed such that the DNN model approximates the optimal resource allocation for feasible cases while for infeasible cases, the trade-off between the objective and the constraint can be achieved, all with low computation time. Our simulation results confirm that near-optimal performance can be achieved for feasible cases, while achieving performance that balances objective and constraint satisfaction in the case of infeasible scenarios, all with low computational overhead.

This paper proposes a resource allocation strategy for multi-channel underlay cognitive radio (CR) systems by means of an ensemble deep learning framework. The transmit power of secondary users (SUs) allocated to each channel is determined to maximize the overall spectral efficiency (SE), whilst meeting the interference constraint on the primary user (PU). To this end, a deep neural network (DNN) structure is developed, in which multiple DNN units are jointly utilized, to obtain the diversity over different DNNs. Our simulation results confirm that the proposed scheme can achieve near-optimal performance with a low computation time of less than 1.5 ms.

Abstract In this study, we investigated a convolutional neural network (CNN)-based framework for the estimation of the best-corrected visual acuity (BCVA) from fundus images. First, we collected 53,318 fundus photographs from the Gyeongsang National University Changwon Hospital, where each fundus photograph is categorized into 11 levels by retrospective medical chart review. Then, we designed 4 BCVA estimation schemes using transfer learning with pre-trained ResNet-18 and EfficientNet-B0 models where both regression and classification-based prediction are taken into account. According to the results of the study, the predicted BCVA by CNN-based schemes is close to the actual value such that 94.37% of prediction accuracy can be achieved when 3 levels of difference can be tolerated during prediction. The mean squared error and $$R^2$$ R 2 score were measured as 0.028 and 0.654, respectively. These results indicate that the BCVA can be predicted accurately for extreme cases, i.e., the level of BCVA is close to either 0.0 or 1.0. Moreover, using the Guided Grad-CAM, we confirmed that the macula and the blood vessel surrounding the macula are mainly utilized in the prediction of BCVA, which validates the rationality of the CNN-based BCVA estimation schemes since the same area is also exploited during the retrospective medical chart review. Finally, we applied the t-distributed stochastic neighbor embedding to examine the characteristics of CNN-based BCVA estimation schemes. The developed BCVA estimation schemes can be employed to obtain the objective measurement of BVCA as well as the medical screening of people with poor access to medical care through smartphone-based fundus imaging.

Abstract This paper investigates the prediction of mobile traffic load based on four variants of recurrent neural networks, which are the simple long short‐term memory (LSTM), stacked LSTM, gated recurrent unit (GRU) and bidirectional LSTM. In the considered schemes, the mobile traffic load of 15 min ahead of time is estimated based on the previous mobile traffic load data. The performance of the proposed scheme is verified using realistic traffic load data collected from the base station located in Kabul city, Afghanistan, which belongs to the SALAAM telecommunication operator during December 2020 and January 2021. Through performance evaluation, the authors confirm that the traffic load can be predicted with high accuracy using considered schemes and the GRU‐based scheme outperforms other schemes in terms of accuracy.

We propose a deep spread multiplexing (DSM) scheme using a DNN-based encoder and decoder and we investigate training procedures for a DNN-based encoder and decoder system. Multiplexing for multiple orthogonal resources is designed with an autoencoder structure, which originates from the deep learning technique. Furthermore, we investigate training methods that can leverage the performance in terms of various aspects such as channel models, training signal-to-noise (SNR) level and noise types. The performance of these factors is evaluated by training the DNN-based encoder and decoder and verified with simulation results.

In magnetic resonance-based wireless power transfer (WPT), any misalignment between resonators degrades the strength of the magnetic coupling, and the use of a relay does not always improve performance if it is misaligned with a transmitter or a receiver. It is therefore necessary to determine the proper activation of a relay depending on the degree of misalignment. In this paper, we analyze the effect of a misaligned relay on two important performance metrics in WPT, namely output power and efficiency. In particular, we derive the achievable output power and efficiency with optimal load resistance based on an equivalent circuit model, and determine whether it is beneficial to utilize the relay for a given lateral and angular misalignment. By simulation and experimental verifications under various environments, we validate the exactness of our analysis, and we also confirm two interesting observations, namely 1) with regard to performance improvement in terms of output power and efficiency, it is better to deactivate the relay if it is seriously misaligned with the other resonators and 2) output power is more vulnerable to the misalignment of a relay than efficiency.

Wireless devices, including smartphones/PDAs and the network equipment that connects them, consume a significant amount of power. The global greenhouse gas emissions produced by the operation of information and communication technology (ICT) infrastructure is comparable to that produced by the aviation industry. This situation is getting worse since billions of wireless devices, such as the Internet of Things devices, are expected to be newly added to the Internet in the coming decade. Consequently, there are multiple research and development projects that aim at addressing this issue, including some IEEE standards intended to reduce the environmental impact of the ICT industry. These standards include energy-efficient communications and networking. Although these emerging techniques have attracted considerable attention and have been studied recently, there are still many open research challenges to be addressed.

In this paper, we propose four transmit power control strategies for the underlay device-to-device (D2D) communications, in which the spectral efficiency (SE) of the D2D communications is maximized while the amount of interference caused to a base station (BS) is kept less than a predefined threshold. To this end, we first propose a centralized power control strategy based on instantaneous and global channel state information (CSI) by formulating a convex optimization problem. Then, three distributed power control strategies are taken into account in which each D2D pair adjusts its transmit power in a distributed manner based on interference price and its local CSI, which significantly reduces the signaling overhead. In the distributed strategies, the interference price can be determined based on (1) the instantaneous local CSI; (2) the statistics of the local CSI (average power); and (3) the number of the D2D pairs without any CSI knowledge. Through extensive computer simulations, we show that the performances of the proposed strategies optimally adjust the transmit power of the D2D communications. Especially, we find that the distributed power control strategies can achieve almost the same SE with the centralized strategy with much lower signaling and control overhead.

Estimating the growth performance in pigs is important in order to achieve a high productivity of pig farming. We herein analyze and verify the machine learning based estimations for the growth performance in swine which includes the daily gain of body weight (DG), feed intake (FI), required growth period for growing/finishing phase (GP), and marketed-pigs per sow per year (MSY), based on the farm specific data and climate, i.e., temperature, humidity, initial age (IA), initial body weight (IBW), number of pigs (NU) and stocking density (SD). The growth data used in our work is collected from 55 pig farms which are located across South Korea for the period between October 2017 and September 2018. In the estimation of growth performance, four machine learning schemes are applied, which are the logistic regression, linear support vector machine (SVM), decision tree, and random forest. Through the evaluation, we confirm that the accuracy of estimation for growth performance can be improved by 28% using machine learning techniques compared to the base line performance which is obtained by the ZeroR classifier. We also find that the accuracy of estimation is heavily dependent on the pre-process of growth data.

Visual acuity (VA) is a measure of the ability to distinguish shapes and details of objects at a given distance and is a measure of the spatial resolution of the visual system. Vision is one of the basic health indicators closely related to a person’s quality of life. It is one of the first basic tests done when an eye disease develops. VA is usually measured by using a Snellen chart or E-chart from a specific distance. However, in some cases, such as the unconsciousness of patients or diseases, i.e., dementia, it can be impossible to measure the VA using such traditional chart-based methodologies. This paper provides a machine learning-based VA measurement methodology that determines VA only based on fundus images. In particular, the levels of VA, conventionally divided into 11 levels, are grouped into four classes and three machine learning algorithms, one SVM model and two CNN models, are combined into an ensemble method in order to predict the corresponding VA level from a fundus image. Based on a performance evaluation conducted using randomly selected 4000 fundus images, we confirm that our ensemble method can estimate with 82.4% of the average accuracy for four classes of VA levels, in which each class of Class 1 to Class 4 identifies the level of VA with 88.5%, 58.8%, 88%, and 94.3%, respectively. To the best of our knowledge, this is the first paper on VA measurements based on fundus images using deep machine learning.

Proliferation of mass market applications and all forms of smart devices have driven the explosive growth of wireless sensor data traffic in recent years. In spite of notable advancements in wireless sensor technologies, energy scarcity due to a limited battery capacity still remains a critical impediment to consumer electronics applications. Thus, the energy harvesting (EH) technology, utilizing extra energy collected from radio frequency (RF) signals, is regarded as a promising solution for addressing the battery problem. This paper presents, analyzes, and discusses a joint sensing and power allocation scheme for cognitive radio in conjunction with EH, in which a secondary transmitter can harvest energy from RF signals transmitted by a primary transmitter during spectrum sensing. We formulate a non-convex optimization problem to find the optimal sensing time and power allocation for maximizing energy efficiency, while satisfying the constraints on the amount of harvested energy and interference at primary receivers. Using nonlinear fractional programming, the original problem can be reformulated into a tractable convex one, and an energy efficient resource allocation algorithm is devised while taking into account RF EH. Simulation results are used to verify the optimality of the proposed scheme, where the secondary network can accomplish maximum energy efficiency.

As multimedia traffic has been increasing and is expected to grow more sharply, various technologies using caches have been attracting lots of attention. As one breakthrough technology to deal with the explosively growing traffic, exclusive OR (XOR)-based index coding has been widely investigated because it can greatly enhance the efficiency of network resource by reducing the number of transmissions. In this paper, we investigate how to apply XOR-based index coding to large-scaled practical streaming systems for video traffic that accounts for more than 70% of total Internet traffic. Contrary to most previous studies that have focused on theoretical analysis of optimal performance or development of optimal index coding schemes, our study proposes a new XOR coding-based video streaming (XC). We also propose a new grouping algorithm for creating XC groups while guaranteeing the complete backward compatibility of XC with existing streaming schemes such as unicast (UC), multicast (MC), and broadcast (BC). The performance of the proposed scheme is analyzed in two steps. First, the behavior of video contents in caches at clients is modeled as a Markov chain, and the steady-state probabilities and caching probabilities for each piece of video content are derived. Based on the probabilities, the performance of the proposed system is then analyzed in terms of the average number of connections that each client requires in order to receive one video content. Our numerical results show that the proposed video streaming scheme using XC can reduce the average number of transmissions by up to 18%, compared to the conventional scheme.

Improving energy efficiency (EE) of mobile communication systems (MCSs) has been considered a key aim in recent years, and has been the subject of intense research interest. One of the simplest yet most powerful ways to increase the EE is to turn off redundant communication entities whose operation does not greatly affect the overall performance of the MCS. In this paper, we propose a novel on–off power control scheme for femto-cell base stations (FBSs) considering cooperative transmission in which multiple FBSs collaborate on the same data transmission. In the proposed scheme, the operation of the redundant FBSs is halted in an adaptive manner. For the proper determination of redundant FBSs with low computational complexity, we propose using the level of contribution (LOC), which specifies the importance of a given FBS in the cooperative transmission. Redundant FBSs are chosen based on their LOC value, and these FBSs are turned off in order to reduce the power consumption of MCSs while minimizing the degradation of the overall throughput of cooperative transmissions. The performance of the proposed scheme is verified through extensive simulations, which shows that near-optimal performance can be achieved without excessive computations.

In this study, we aim to validate the use of Artificial Intelligence of Things (AIoT) for deep learning-based detection of piglet crushing incidents by sows in pig farms, a leading cause of piglet mortality. To achieve this, we developed a deep neural network based on the You Only Look Once (YOLO) model, which was trained to detect piglet crushing events in real-time. We then optimized the weights of the YOLO model using TensorFlow Lite, TensorRT, and edge TPU to ensure efficient operation on AIoT devices with limited computational resources. Finally, we compared the performance of the deep learning-based piglet crushing detection algorithm on various single-board computers (SBCs), including the Jetson Nano, Raspberry Pi, and Coral Dev Board, to validate its suitability for real-world applications. [ABSTRACT FROM AUTHOR]