첨단 무선 통신망을 위한 AI 기반 무선 자원 관리

Abstract

Nowadays, wireless communications systems have grown significantly. Therefore, using the radio frequency (RF) spectrum is increasing more dramatically in order to meet the growth in applications requiring broadband. Because of the spectrum scarcity, new spectrum management models are being developed to opportunistically utilize the dynamic spectrum access. Currently, the cognitive radio network (CRN) was developed and is considered one of the most promising technologies for improving spectrum efficiency. Cognitive radio (CR) was recognized as an enabling technology to mitigate the abuse of scarce RF spectrum in which dynamic spectrum access was proposed to share the available spectrum through opportunistic usage of the frequency bands by secondary operators without interfering with the primary networks. When CRN is installed, it enables secondary networks to perform the following tasks: spectrum sensing, spectrum management, and mobility management. CR technology offers the opportunity to optimize spectrum access, assists in preventing interference, and adapts to instant spectrum slot availability from the unused spectrum pool.

Along with rapid developments of mobile applications as well as expanding network infrastructure (transmission lines, terminal equipment, and base stations), efficient energy management is becoming an issue that deserves special attention. Efficient energy management helps to overcome the bottleneck of wireless network applications operating under battery and energy constraints. It not only helps to reduce a device's dependence on battery power and power consumption, but also provides a continuous power source for the long-term operation of devices on the network. As a result, wireless communications powered by external harvested energy and the simultaneous wireless information and power transfer (SWIPT) transmission have become promising techniques to solve the energy-constrained problem. Regarding external harvested energy, radio frequency-energy harvesting (RF-EH) is a potential solution for energy-constrained issue in wireless networking, where the wireless devices can harvest energy from ambient RF signals. Along with RF-EH, non-RF energy resources (e.g., solar, wind, etc.) can also provide perpetual energy and higher power density for rechargeable batteries of wireless users. Regarding the SWIPT system, both information and energy of the common transmit signal are transmitted to the receivers. Therefore, the received signal can be used for energy harvesting (EH) and information decoding (ID).

In addition, artificial intelligence (AI), which is defined as any process or device that perceives its environment and takes actions that maximize the chances of success for some predefined goal, is a feasible solution for the emerging complex communication system design. The recent advances in reinforcement learning (RL) and deep learning (DL) hold significant promise for solving very complex problems considered intractable until now. It is now appropriate to apply AI technology to advanced wireless communication networks (WCNs) to tackle optimized complicated decision making, physical layer design, network management and resource optimization tasks in such networks. In the study of wireless technologies and communication systems, AI will be a powerful tool and hot research topic with many potential application areas, e.g., channel modeling, wireless signal processing, and resource management. Motivated by the aforementioned survey, this dissertation will focus on these remaining issues about applying AI to radio resource management for advanced WCNs, such as CRN and SWIPT, as follows:

Firstly, in CRN, secondary users (SUs) are able to sense the absence of primary users (PUs) in the spectrum. Then, SUs use this information to opportunistically access the licensed spectrum. In this work, we utilize a software-defined radio testbed of energy detection (ED)-based spectrum sensing. The testbed was built based on the GNU’s Not Unix (GNU) Radio software platform and Universal Software Radio Peripheral (USRP) National Instruments 2900 devices. In this case, a new block of energy detection is developed by using an out-of-tree module from GNU Radio. To successfully integrate CR into the cloud computing paradigm, we also implement cloud computing-based spectrum sensing by utilizing a cloud server with ThingSpeak, such that we can store, process, and share the sensing information more efficiently in a centralized way in the cloud server. In addition, we also present an implementation of real-time video transmission with spectrum-sensing among two USRP devices. In this work, spectrum-sensing is implemented at both transmitter and receiver. The transmitter senses the channel, and if the channel is free, a video signal (which could be a real-time signal from a video file) will be modulated and processed by GNU Radio and transmitted using a USRP. A USRP receiver also senses the channel, but in contrast, if the channel is busy, the signal is demodulated to reproduce the transmitted video signal. These works brings in several challenges, like spectrum-sensing in the devices' environment, and packets getting lost or corrupted over the air.

Secondly, although a CRN is a novel solution that promises to solve the spectrum scarcity problem and enhance spectrum utilization, unsecured CRN can easily be manipulated in order to attack legacy users on the communication channel. As a result, the network's performance significantly degrades. Therefore, communication channel security is an important issue that needs to be addressed in a CRN. In this work, we focus on improving the security of multi-channel communication in a CRN, while various jammers try to access channels of interest to prevent SUs from using them. By using game-theoretic concepts and by defining states, actions, and players’ rewards, we propose game-based schemes that find the best channel for the SUs in order to avoid jammer's attacks on communication channels. Accordingly, the problem is finding the optimal channel to maximize the long-term reward of the SU where communication channels are not used by the PUs and are not jammed by attackers. In addition, the idea of transfer learning might be applied to the problem under consideration, and thus, a transfer Game-Actor-Critic (TGACT) scheme is proposed, which uses the transferred knowledge in a double-game period to accelerate the learning process and provide performance improvement in channel selection. The simulation results show that the proposed schemes are quite resistant to jammer attacks, and achieve better performance compared to other channel selection schemes.

Thirdly, the SWIPT systems can supply efficiently throughput and energy, have emerged as a potential research area in fifth-generation (5G) system. In this work, We investigate the SWIPT system with multi-user, single-input single-output (SISO) system. First, we solve the transmit power optimization problem, which provides the optimal strategy for getting minimum power while satisfying sufficient signal-to-interference-plus-noise ratio (SINR) and harvested energy requirements to ensure receiver circuits work in SWIPT systems where receivers are equipped with a power-splitting (PS) structure. Although optimization algorithms are able to achieve relatively high performance, they often entail a significant number of iterations, which raises many issues in computation costs and time for real-time applications. Therefore, we aim at providing a DL-based approach, which is a promising solution to address this challenging issue. DL architectures used include a type of Deep Neural Network (DNN): the Feed-Forward Neural Network (FFNN) and three types of Recurrent Neural Network (RNN): the Layer Recurrent Network (LRN), the Nonlinear AutoRegressive network with eXogenous inputs (NARX), and Long Short-Term Memory (LSTM). Through simulations, we show that the DL-based approach can approximate a complex optimization algorithm that optimizes transmit power in SWIPT systems with much less computation time.

Then, the demand for spectral and energy efficiency has significantly been increased along with new breakthroughs in programmable meta-material techniques. The integration of an intelligent reflecting surface (IRS) into the SWIPT systems has attracted much attention from operators in advanced WCNs such as 5G and sixth-generation (6G) networks. In addition, an IRS-assisted SWIPT system faces many security risks that can easily be compromised by eavesdroppers. In this work, we investigate the physical-layer secure and transmission optimization problem in an IRS-assisted SWIPT system where a PS scheme is installed in the user equipment (UE). In particular, our purpose is to maximize the system secrecy rate by jointly finding optimal solutions for transmitter power, PS factor of UE, and phase shifts matrix of IRS under the required minimum harvested energy and maximum transmitter power. We propose the alternating optimization (AO)-based scheme to obtain optimal solutions. The proposed AO-based scheme can effectively solve both convex and non-convex problems; however, applying them in practice still poses some difficulties due to the complexity and long computation time. This is because many mathematical transformations are used and the optimal solution needs a number of iterations to achieve convergence. Therefore, we also propose 5 types of data and DNN structures to potentially achieve efficiency in computations by using a DL-based approach. The simulation results indicate that the proposed IRS scheme provides an improvement in terms of the average secrecy rate (ASR) by up to 38.91% when the number of reflecting elements is high (30 elements) compared to a scheme without an IRS. We also observe that the DL-based approach not only provides similar performance to the AO-based scheme but it also significantly reduces computation time.

Consequently, we end up this dissertation by summarizing its main contributions and opening a new door for RL, DL techniques and AI algorithms in future wireless networks.