서브메뉴
검색
Fiber-Wireless Integration with Enhanced Adaptability for Next Generation Radio Access Networks
Fiber-Wireless Integration with Enhanced Adaptability for Next Generation Radio Access Networks
Detailed Information
- 자료유형
- 학위논문 서양
- 최종처리일시
- 20260202105538
- ISBN
- 9798263394448
- DDC
- 621.3822
- 저자명
- Yao, Shuang.
- 서명/저자
- Fiber-Wireless Integration with Enhanced Adaptability for Next Generation Radio Access Networks
- 발행사항
- [Sl] : Georgia Institute of Technology, 2022
- 발행사항
- Ann Arbor : ProQuest Dissertations & Theses, 2022
- 형태사항
- 164 p
- 주기사항
- Source: Dissertations Abstracts International, Volume: 87-05, Section: B.
- 주기사항
- Advisor: Ma, Xiaoli.
- 학위논문주기
- Thesis (Ph.D.)--Georgia Institute of Technology, 2022.
- 초록/해제
- 요약In next generation radio access networks (RANs), there will be no "one-size-fits-all" solution. A wide range of applications have to be supported and they have been classified into three usage scenarios with diverged requirements, enhanced mobile broadband (eMBB) for high data rates, ultra-reliable and low-latency communications (URLLC) for low latency and high reliability, and massive machine type communications (mMTC) for high connection density. Diverse deployment options are expected, with both newly installed RANs and the existing infrastructure utilized for service delivery. A new RAN architecture has emerged, where eight function split options are defined to redistribute signal processing functions between central unit (CU), distributed unit (DU) and remote radio unit (RRU), and flexible function split is supported as well. At the physical layer, frequency range 1 (FR1) and FR2 have been standardized and scalable numerology has been adopted. To this end, RANs with enhanced adaptability are highly desirable.Fiber-wireless integration combines the high bandwidth of optical fibers with the ubiquitous coverage of wireless links. It has been successfully employed in the fourth generation (4G) mobile communications and will continue to underpin next generation RANs. This dissertation focuses on advanced techniques applied to fiber-wireless integration for adaptability enhancement in RANs. Considering different characteristics of analog radio-over-fiber (RoF) and digital RoF systems, techniques applied to these two systems are investigated separately. In addition, a hybrid system that supports co-delivery of analog RoF and digital RoF signals is also studied, so that the benefits of each format can be utilized to carry different services.Analog RoF preserves the waveform of the radio frequency (RF) signal in the optical domain and offers the simplest RRU configuration. The transparency nature indicates that the resources in the optical domain and RF domain have to be managed cooperatively. The quality of transmission (QoT) estimation, which predicts the received signal quality before service provision and enables resource allocation to adapt to channel conditions, also needs to take the impairments in both fiber and wireless links into account. This poses a challenge to accurate QoT estimation that is critical to low-margin network operation. In this dissertation, artificial neural networks (ANNs) are employed to achieve high accuracy QoT estimation, whose expressiveness can capture the complicated channel model in analog RoF systems. An experimental demonstration is completed in a fiber-millimeter wave (mmWave) testbed where two user equipments (UEs) access the same frequency resources through power-domain non-orthogonal multiple access (PD NOMA). The ANN presents high prediction accuracy even if there is interference between UEs. Furthermore, the data efficiency of ANN-based QoT estimation is promoted via active learning. A data selection framework is proposed that selects unlabeled training samples with high model uncertainty. It can be incorporated into the original ANN training with minimal modifications. The proposed framework is experimentally verified and the number of training samples can be reduced without sacrificing model accuracy. An analysis is also conducted that inspects the distribution of the training samples selected by active learning and interprets the improvement of active learning.Digital RoF transmits the digitized RF signal over the fiber. The adoption of massive multiple-input multiple-output (MIMO) and large channel bandwidth demands high-capacity digital RoF systems. This dissertation looks into issues that would occur as digital RoF systems migrate to high-speed operation. In intensity-modulation and direct-detection (IM/DD) schemes, the bandwidth limitation becomes increasingly severe with the increase of symbol rate and digital signal processing (DSP) based channel equalization is compulsory. Least-mean squares (LMS) algorithm is widely used in equalizer training due to its computational simplicity, but it suffers from slow convergence if there is severe channel distortion. The convergence issue is approached by transmitter-side spectral shaping. Specifically, a first-order Markov chain (MC) is utilized to generate a sequence with correlated samples. The comparison between the traditional sequence with independent and identically distributed (i.i.d.) samples is made through both simulations and experiments. A faster rate of convergence is observed in both tap coefficients and mean-squared error (MSE) compared with the i.i.d. sequence. It gives rise to lower pre-forward-error-correction (pre-FEC) bit error rate (BER) under a fixed sequence length when both MC sequence and i.i.d. sequence are used for training. As a result, significant reduction in training sequence length is attained. Moreover, adjusting the hyperparameter of the first-order MC can change the spectral shaping of the MC sequence, which can be used to adapt to various system bandwidths.Coherent communication systems that can further increase the data rate of mobile fronthaul are also studied and the laser phase noise tolerance is enhanced by probabilistic shaping (PS). An angular distance directed (ADD) distribution is proposed where constellation points in the quadrature amplitude modulation (QAM) with larger angular distances are assigned higher probabilities of transmission. This is based on the observation that constellation points with larger angular distances are more robust against phase noise. The ADD distribution is added on top of the conventional Maxwell-Boltzmann (MB) distribution to get an MB/ADD distribution that attains improved laser phase noise tolerance and SNR tolerance at the same time. The performance of the MB/ADD distribution is investigated with simulations and experiments and it achieves lower pre-FEC BER and higher generalized mutual information (GMI) under large laser phase noise, compared with uniform distribution and the MB distribution. The MB/ADD distribution thus relaxes the requirement on laser linewidth and enables low-cost distributed feedback (DFB) lasers to be used for coherent communications. Besides, both MB part and ADD part in the MB/DD distribution has its own shaping parameter, and they can be agilely selected according to channel SNR and laser linewidth.
- 일반주제명
- Radio frequency
- 일반주제명
- Virtual reality
- 일반주제명
- Neural networks
- 일반주제명
- Signal processing
- 일반주제명
- Electrical engineering
- 일반주제명
- Information technology
- 일반주제명
- Optics
- 기본자료저록
- Dissertations Abstracts International. 87-05B.
- 전자적 위치 및 접속
- 로그인 후 원문을 볼 수 있습니다.
MARC
008260126s2022 us c eng d■001000017360505
■00520260202105538
■006m o d
■007cr#unu||||||||
■020 ▼a9798263394448
■035 ▼a(MiAaPQ)AAI32314911
■035 ▼a(MiAaPQ)GeorgiaTech66626
■040 ▼aMiAaPQ▼cMiAaPQ
■0820 ▼a621.3822
■1001 ▼aYao, Shuang.
■24510▼aFiber-Wireless Integration with Enhanced Adaptability for Next Generation Radio Access Networks
■260 ▼a[Sl]▼bGeorgia Institute of Technology▼c2022
■260 1▼aAnn Arbor▼bProQuest Dissertations & Theses▼c2022
■300 ▼a164 p
■500 ▼aSource: Dissertations Abstracts International, Volume: 87-05, Section: B.
■500 ▼aAdvisor: Ma, Xiaoli.
■5021 ▼aThesis (Ph.D.)--Georgia Institute of Technology, 2022.
■520 ▼aIn next generation radio access networks (RANs), there will be no "one-size-fits-all" solution. A wide range of applications have to be supported and they have been classified into three usage scenarios with diverged requirements, enhanced mobile broadband (eMBB) for high data rates, ultra-reliable and low-latency communications (URLLC) for low latency and high reliability, and massive machine type communications (mMTC) for high connection density. Diverse deployment options are expected, with both newly installed RANs and the existing infrastructure utilized for service delivery. A new RAN architecture has emerged, where eight function split options are defined to redistribute signal processing functions between central unit (CU), distributed unit (DU) and remote radio unit (RRU), and flexible function split is supported as well. At the physical layer, frequency range 1 (FR1) and FR2 have been standardized and scalable numerology has been adopted. To this end, RANs with enhanced adaptability are highly desirable.Fiber-wireless integration combines the high bandwidth of optical fibers with the ubiquitous coverage of wireless links. It has been successfully employed in the fourth generation (4G) mobile communications and will continue to underpin next generation RANs. This dissertation focuses on advanced techniques applied to fiber-wireless integration for adaptability enhancement in RANs. Considering different characteristics of analog radio-over-fiber (RoF) and digital RoF systems, techniques applied to these two systems are investigated separately. In addition, a hybrid system that supports co-delivery of analog RoF and digital RoF signals is also studied, so that the benefits of each format can be utilized to carry different services.Analog RoF preserves the waveform of the radio frequency (RF) signal in the optical domain and offers the simplest RRU configuration. The transparency nature indicates that the resources in the optical domain and RF domain have to be managed cooperatively. The quality of transmission (QoT) estimation, which predicts the received signal quality before service provision and enables resource allocation to adapt to channel conditions, also needs to take the impairments in both fiber and wireless links into account. This poses a challenge to accurate QoT estimation that is critical to low-margin network operation. In this dissertation, artificial neural networks (ANNs) are employed to achieve high accuracy QoT estimation, whose expressiveness can capture the complicated channel model in analog RoF systems. An experimental demonstration is completed in a fiber-millimeter wave (mmWave) testbed where two user equipments (UEs) access the same frequency resources through power-domain non-orthogonal multiple access (PD NOMA). The ANN presents high prediction accuracy even if there is interference between UEs. Furthermore, the data efficiency of ANN-based QoT estimation is promoted via active learning. A data selection framework is proposed that selects unlabeled training samples with high model uncertainty. It can be incorporated into the original ANN training with minimal modifications. The proposed framework is experimentally verified and the number of training samples can be reduced without sacrificing model accuracy. An analysis is also conducted that inspects the distribution of the training samples selected by active learning and interprets the improvement of active learning.Digital RoF transmits the digitized RF signal over the fiber. The adoption of massive multiple-input multiple-output (MIMO) and large channel bandwidth demands high-capacity digital RoF systems. This dissertation looks into issues that would occur as digital RoF systems migrate to high-speed operation. In intensity-modulation and direct-detection (IM/DD) schemes, the bandwidth limitation becomes increasingly severe with the increase of symbol rate and digital signal processing (DSP) based channel equalization is compulsory. Least-mean squares (LMS) algorithm is widely used in equalizer training due to its computational simplicity, but it suffers from slow convergence if there is severe channel distortion. The convergence issue is approached by transmitter-side spectral shaping. Specifically, a first-order Markov chain (MC) is utilized to generate a sequence with correlated samples. The comparison between the traditional sequence with independent and identically distributed (i.i.d.) samples is made through both simulations and experiments. A faster rate of convergence is observed in both tap coefficients and mean-squared error (MSE) compared with the i.i.d. sequence. It gives rise to lower pre-forward-error-correction (pre-FEC) bit error rate (BER) under a fixed sequence length when both MC sequence and i.i.d. sequence are used for training. As a result, significant reduction in training sequence length is attained. Moreover, adjusting the hyperparameter of the first-order MC can change the spectral shaping of the MC sequence, which can be used to adapt to various system bandwidths.Coherent communication systems that can further increase the data rate of mobile fronthaul are also studied and the laser phase noise tolerance is enhanced by probabilistic shaping (PS). An angular distance directed (ADD) distribution is proposed where constellation points in the quadrature amplitude modulation (QAM) with larger angular distances are assigned higher probabilities of transmission. This is based on the observation that constellation points with larger angular distances are more robust against phase noise. The ADD distribution is added on top of the conventional Maxwell-Boltzmann (MB) distribution to get an MB/ADD distribution that attains improved laser phase noise tolerance and SNR tolerance at the same time. The performance of the MB/ADD distribution is investigated with simulations and experiments and it achieves lower pre-FEC BER and higher generalized mutual information (GMI) under large laser phase noise, compared with uniform distribution and the MB distribution. The MB/ADD distribution thus relaxes the requirement on laser linewidth and enables low-cost distributed feedback (DFB) lasers to be used for coherent communications. Besides, both MB part and ADD part in the MB/DD distribution has its own shaping parameter, and they can be agilely selected according to channel SNR and laser linewidth.
■590 ▼aSchool code: 0078.
■650 4▼aWave division multiplexing
■650 4▼aRadio frequency
■650 4▼aVirtual reality
■650 4▼aNeural networks
■650 4▼aSignal processing
■650 4▼aElectrical engineering
■650 4▼aInformation technology
■650 4▼aOptics
■690 ▼a0800
■690 ▼a0544
■690 ▼a0489
■690 ▼a0752
■71020▼aGeorgia Institute of Technology.
■7730 ▼tDissertations Abstracts International▼g87-05B.
■790 ▼a0078
■791 ▼aPh.D.
■792 ▼a2022
■793 ▼aEnglish
■85640▼uhttp://www.riss.kr/pdu/ddodLink.do?id=T17360505▼nKERIS▼z이 자료의 원문은 한국교육학술정보원에서 제공합니다.
Preview
Export
ChatGPT Discussion
AI Recommended Related Books
Подробнее информация.
- Бронирование
- не существует
- моя папка
- Первый запрос зрения
- Non-Book Loan Application
- Nighttime Book Loan Application
Available after logging in.


