Radio access network (RAN) in 5G is expected to satisfy the stringent delay requirements of a variety of applications. The packet scheduler plays an important role by allocating spectrum resources to user equipments (UEs) at each transmit time interval (TTI). In this paper, we show that optimal scheduling is a challenging combinatorial optimization problem, which is hard to solve within the channel coherence time with conventional optimization methods. Rule-based scheduling methods, on the other hand, are hard to adapt to the time-varying wireless channel conditions and various data request patterns of UEs. Recently, integrating artificial intelligence (AI) into wireless networks has drawn great interest from both academia and industry. In this paper, we incorporate deep reinforcement learning (DRL) into the design of cellular packet scheduling. A delay-aware cell traffic scheduling algorithm is developed to map the observed system state to scheduling decision. Due to the huge state space, a recurrent neural network (RNN) is utilized to approximate the optimal action-policy function. Different from conventional rule-based scheduling methods, the proposed scheme can learn from the interactions with the environment and adaptively choosing the best scheduling decision at each TTI. Simulation results show that the DRL-based packet scheduling can achieve the lowest average delay compared with several conventional approaches. Meanwhile, the UEs' average queue lengths can also be significantly reduced. The developed method also exhibits great potential in real-time scheduling in delay-sensitive scenarios.

Universal filtered multi-carrier (UFMC) has been widely researched for 5G communication systems. In this paper, we propose a new UFMC receiver model and signaling structure to support asynchronous transmissions. Since the UFMC transmitter contains sub-band filters, we adopt equivalent filters at the receiver to separate target user signal over asynchronous multi-user transmissions. Moreover, we introduce a coarse timing synchronization algorithm in the presence of multiple timing offsets (TOs) over the complex additive white Gaussian noise (AWGN). Finally, simulation results show that the performance of the proposed system with asynchronous users is better than conventional 4G uplink system with synchronous users.

Centralized radio access networks (RANs) based on radio-over-fiber (RoF) techniques can simplify the remote access units (RAUs) and maximize the resource sharing among services and operators, thus improving the system efficiency and reducing the operational cost. A scheme of resource sharing is proposed for centralized RANs based on millimeter-wave (MMW) RoF system. Both RoF infrastructure resources and microwavephotonics resources are shared by multiple operators in the heterogeneous system. High-throughput services carried on both V-band (60GHz) and W-band (100GHz) are delivered to RAUs through flexible configuration at the central office (CO). This proposed multi-service centralized RAN with shared infrastructure is successfully demonstrated via both simulations and experiments.

A theoretical and experimental investigation of the nonlinear behaviors of intra- and inter-band cross-modulation (XM) in a multiband radio-over-fiber (RoF) system is presented for the first time. A closed-form theoretical analysis is proposed to characterize and validate the data XM induced by the nonlinearity of optical Mach-Zehnder modulator (MZM). Nonlinear behaviors of the intra-band XM between the in-phase and quadrature (I/Q) components of a single-band vector signal, as well as the inter-band XM between multiband combinations of scalar and vector signals are investigated, and consistent agreement has been verified between the theory and experiments. It was confirmed that scalar signals only suffer from inter-band XM; whereas vector signals suffer from both intra- and inter-band XMs. Intra-band XM induces constellation compression to vector signals; while inter-band XM induces eye diagram degradation to scalar signals, as well as nonlinear constellation scaling to vector signals. Moreover, QPSK shows superior tolerance to MZM nonlinearity than higher order QAM formats due to it intrinsic resistance to intra-band XM.

Small cells have become an integral component in meeting the increased demand for cellular network capacity. Cloud radio access networks (C-RAN) have been proposed as an effective means to harness the capacity benefits of small cells at reduced capital and operational expenses. With the baseband units (BBUs) separated from the radio access units (RAUs) and moved to the cloud for centralized processing, the backhaul between BBUs and RAUs forms a key component of any C-RAN. In this work, we argue that a one-one mapping of BBUs to RAUs is highly sub-optimal, thereby calling for a functional decoupling of the BBU pool from the RAUs. Further, the backhaul architecture must be made re-configurable to allow the mapping between BBUs and RAUs to be flexible and changed dynamically so as to not just optimize RAN performance but also energy consumption in the BBU pool. Towards this end, we design and implement the first OFDMA-based C-RAN test-bed with a reconfigurable backhaul that allows 4 BBUs to connect flexibly with 4 RAUs using radio-over-fiber technology. We demonstrate the feasibility of our system over a 10 km separation between the BBU pool and RAUs. Further, real world experiments with commercial off-the-shelf WiMAX clients reveal the performance benefits of our reconfigurable backhaul in catering effectively to heterogeneous user (static and mobile clients) and traffic profiles, while also delivering energy benefits in the BBU pool.

Carrier Ethernet is becoming the dominating backhaul network due to its flexibility, scalability, interoperability and low-cost. Carrier Ethernet networks (CENs) are inherently high bandwidth-delay product (BDP) networks, hence Traditional TCP, and similar variants, make undesirable choices of transport protocol mainly because of the aggressive multiplicative-decrease algorithm. The Ethernet-Services Transport Protocol (ESTP) has proven relieve the effects of the decrease algorithm by dynamically adjusting the transmission rate according to the level of congestion estimated in the network. It should be emphasized that most protocols detect congestion, but do not estimate the level of congestion. Also ESTP incorporates QoS information to further improve performance. In this work, a steady-state throughput analytical model of ESTP is derived, an experimental testbed is built and simulation results are obtained. All validation methods show good agreement.

A radio-over-fiber system can provide a robust backbone network for the deployment of multi-gigabit wireless local area networks. By transmitting 2×2, MIMO signals over two polarizations in an optical fiber, we developed a new radio-over-fiber system to support MIMO-based wireless communications to improve the signal diversity or multiplexity. Two-Gb/s millimeter-wave signals were successfully transmitted through a 10-km optical fiber and a wireless channel with the equivalent spectral efficiency of 2bits/Hz using MIMO spatial multiplexing.

Delivering affordable high-quality E-health care services requires effective telecommunication and multimedia technologies in conjunction with medical expertise. In this paper, we propose and demonstrate a telecommunication and networking architecture for implementing next generation Telemedicine and Telehealth systems. Our integrated optical-wireless based network provides super broadband, ultra low-latency connectivity for voice, video, image and data across various telemedicine modalities to facilitate real-time and near real-time communication of remote health care information.

The metro Ethernet network (MEN) expands the advantages of Ethernet to cover areas wider than LAN. Ethernet Services were created to provide multiple services to this carrier-grade network that is rapidly growing popularity in North America [21], Europe [22], and Asia [23]. Techniques to police the network layer for MENs are very mature; but the transport layer relies on different flavors of TCP that do not get the most of Ethernet services. We propose a transport layer protocol that can replace previous versions of TCP to better suit metro Ethernet service policies, such as those established for specified service level agreements (SLAs). Ethernet service's traffic and performance parameters include committed information rate (CIR), excess information rate (EIR), and guaranteed packet loss rate. The purpose of this transport protocol is to integrate these parameters, which are applied at the networking layer, into the proposed transport protocol to make a smarter controller.

This paper aims to enhance the transport services of Ethernet MANs by detecting congestion more accurately and responding to packet losses more adaptively. Traditional TCP was designed for best-effort networks, and cannot utilize available bandwidth efficiently in a broadband link. Most of the previous research addressed this problem by modifying the additive increase multiplicative decrease (AIMD) algorithm of TCP. In this paper, a novel resilient transport control scheme is proposed for Ethernet services that utilize a less aggressive congestion window reduction algorithm when a non-congested environment is detected. A stochastic model for packet loss is proposed, in which packet losses are expressed as a random function of congestion and random loss present in the network Transport layer congestion control is stated as a hypothesis test process; making decisions based on multiple ACK feedbacks. The new method can keep false-alarm probability under control and therefore it increases throughput significantly.

This paper aims to enhance the transport services of Ethernet MANs and WANs by addressing the throughput optimization issue at the edge of the network. Ethernet services convert Ethernet from a best effort technology to a service level agreement (SLA) driven carrier-grade technology, extending the simplicity and flexibility of Ethernet beyond the LAN to the MAN/WAN. Ethernet virtual private line (EVPL) provides a point-to-point virtual leased line that takes advantage of Ethernet's statistical multiplexing to share resources among multiple customers. Although current bandwidth allocation schemes are able to reassign idle capacity and provide QoS at link layer, the end-to-end throughput in the application layer is not optimized. The performance of TCP over SLA driven Ethernet services is not adequate, because the mechanisms of TCP are designed for best effort networks. In this paper, a novel SLA-aware transport protocol is proposed to increase the end- to-end throughput for wide area Ethernet services. The performance of the proposed protocol is compared with traditional TCP through theoretical analysis and simulations.

The optical packet-switching network is considered to be one of the most promising solutions for end-to-end delivery of high-bitrate data, video, and voice signals across optical networks of the future. Optical label switching (OLS) technology incurs simpler extraction and processing of the labels so that the optical packets can be routed with low latency to the destinations. We have developed several key enabling technologies for integrated optical networks, including optical label generation, label swapping, optical buffering, clock recovery, and wavelength conversion. We have designed and experimentally demonstrated these enabling techniques that can provide efficient broadband services in future optical networks.

Delta compression (or differencing) algorithms take advantage of the statistical correlations between different files or packets so that common sequences between two files can be encoded using a copy reference. Delta compression forms the core of fast and efficient content download. In this paper, we present for the first time, a theoretical framework for delta compression based on information theory and Markov models, including insights into the compression bounds. We also simulated and implemented a generic delta compression scheme and demonstrated its real-time and non real-time performance by applying the scheme on binary data. It will be shown that our delta compression scheme is able to improve the real-time performance of wireless Internet download by up to 4 times.

Wavelength division multiplexing passive optical networks (WDM PONs) can dynamically offer each end user a unique optical wavelength for data transmission as well as the possibility of wavelength reuse and aggregation, thereby ensuring scalability in bandwidth assignment. In this paper, we propose a new byte size clock (BSC) reservation MAC scheme that not only arbitrates upstream transmission and prevents optical collisions, but also varies bandwidth according to demand and priority, reduces request delay using pre-allocation and delta compression, and handles the addition/reconfiguration of network nodes efficiently. The new access scheme exploits both WDM and TDM to cater for both light and heavy bandwidth requirements and supports both Ethernet and ATM packets without segmenting or aggregating them. Our proposed protocol is not only backward compatible with APON and EPON, but also provides a better utilization of the access link in terms of the throughput and delay. In addition, the amount of pre-allocated bandwidth can be minimized using delta compression, which in turns reduces the latency due to the request and grant mechanism. We analyzed, evaluated, and simulated the performance and practicality of the proposed scheme.

This work proposes algorithms to achieve full mesh connectivity in 4-fiber WDM self-healing rings (SHRs). The number of wavelengths required is calculated for both odd and even number of nodes on the ring and an optimal wavelength assignment algorithm is presented. This algorithm uses a simple matrix approach for calculating the wavelength assignment between nodes on the ring so as to achieve full mesh connectivity while avoiding any possible violation of the color clash constraint.