2nd International Conference on 5G for Ubiquitous Connectivity

Deep learning makes people enjoy a more convenient as well as smarter mobile life. It is an interesting yet much challenging topic to enable efficient mobile deep learning applications. Traditional approach to tackle this challenge is to employ cloud computing by offloading the computation tasks to remote servers, which has weaknesses of high bandwidth requirements and transmission latency. In this paper, we propose to enable collaborative inference among local mobile devices. Instead of sending deep learning inference tasks to cloud, we let mobile devices collaboratively share the computation workloads. This is based on an important observation that batching inference tasks on GPUs can accelerate the processing speed. To achieve efficient collaboration, we design an algorithm based on partial swarm optimization (PSO) that is a versatile population-based stochastic optimization technique. Moreover, extensive simulations are conducted to evaluate the performance of the designed algorithm. The simulation results show that the collaborative inference scheme can reduce global dealing time in given field compared with offloading tasks to cloud.

The evolution of the LTE standard with Release-14 has introduced several changes in the physical layer with a goal of catering to the needs of Intelligent Transportation Systems (ITS). Some of the changes include the use of SC-FDMA, a modified pilot pattern and improved scheduling. Typical vehicular environments are characterized by a multipath channel that is temporally varying. Such a channel affects the performance of an underlying SC-FDMA system by introducing Inter-Carrier Interference (ICI). Additionally, the channel characteristics are dynamic, resulting in heterogeneous channel conditions with varying degrees of mobility. Channel estimation and equalization schemes are indispensable in overcoming the adverse ramifications of the channel by precisely parameterizing the channel through computationally feasible means. Here, an optimized channel characterization scheme that involves the Rake-Matching Pursuit (RMP) algorithm for the estimation of the so called doubly selective channels and the Least Squares (LS) scheme in low mobility conditions is suggested. A simple and low complex cognitive framework based on the Index of Dispersion is also proposed to switch between the two schemes based on channel conditions. By virtue of its merits, the proposed channel estimation scheme provides good results in heterogeneous channel conditions and with a complexity that remains relevant for consumer hardware. Finally, an evaluation platform in Matlab^® that is LTE standard compliant is made available to the research community. (LTE Evaluation Platform: https://​git.​nt.​uni-saarland.​de/​HiMo/​SimTool_​LTE.) It includes the proposed schemes for channel estimation along with the cognitive framework.

In this paper, the power allocation (PA) for maximizing sum rate in downlink cloud radio access networks based on non-orthogonal multiple access (C-RAN-NOMA) is explored. Specifically, we consider a downlink C-RAN which includes two paired NOMA-users served by several uniformly distributed remote radio heads (RRHs) simultaneously. Based on the C-RAN-NOMA model, a linear programming (LP) algorithm is proposed to maximize the sum rate of downlink, and the corresponding optimal PA scheme is developed for C-RAN-NOMA. Simulation results illustrate that the developed power allocation scheme can achieve higher sum rate than the conventional equal power allocation scheme, and has the same sum rate as the exhaustive search algorithm but with lower complexity. Thus, the effectiveness of the proposed scheme and the corresponding algorithm are well verified.

In this paper, we develop an energy-efficient optimization scheme for uplink distributed antenna system (DAS) with Device-to-Device (D2D) communication. Firstly, we establish the DAS model with D2D communication. Then, the optimization problem for energy efficiency (EE) maximization with the constraints of cellular user and D2D rates are formulated. Based on the pseudo-concave of objective function in optimization problem, we propose an optimal power allocation algorithm with gradient descent method and Armijo method to obtain the optimal solution of the optimization problem. The simulation results demonstrate the effectiveness of our proposed scheme, and superior EE performance is achieved.

Due to numerous base stations (BSs) being randomly deployed in ultra-dense networks (UDNs), the co-tier interference and cross-tier interference become more and more serious. We propose an interference management scheme based on joint clustering, subchannel allocation and successive cancellation among users for UDN taking all types of interference into consideration. To reduce the co-tier interference, we design a clustering algorithm based on interference graph to divide femto base stations (FBSs) and femto user equipments (FUEs) into FBS clusters and FUE groups respectively. Then a subchannel allocation algorithm is presented for FBS clusters to avoid the cross-tier interference. Moreover, a successive interference cancellation (SIC) detection algorithm is used to abate the interference among users in the same FUE group. Our analysis and simulations show that the proposed scheme outperforms other schemes in terms of system capacity and spectral efficiency in several scenarios. The results also verify that the proposed scheme can not only effectively mitigate the interference, but also have advantages in meeting the requirements of users and improving the network performance.

Mobile Edge Cloud Prototype (MEC) allows resource-constraint mobile device to execute computation intensive and delay sensitive applications (i.e., Augmented Reality, Healthcare, Virtual Reality and so on) in a collaborative manner. The offloading system in the MEC is a technique which divides the application execution into local execution and cloud execution in order to augment the user quality of experience (QOS). In this paper, we are formulating an application partitioning and task scheduling problem for delay sensitive healthcare application. To cope up with the aforementioned problem we have proposed a novel Dynamic Aware Application Partitioning Task Scheduling Algorithm (DAPTS) which determine the following phases: (i) partition the application into local and remote execution via static analysis and profiling technology, (ii) schedule a local task on the mobile device, (iii) schedule the cloud tasks via the wireless channel band, (iv) schedule the offloaded tasks on the cloud resources. Simulation results show that propose DAPTS outperforms as compared to baseline approaches in the context of average response time of the application.

Ultra-dense network (UDN) downlink interference is researched in this paper, where some neighboring base stations significantly interfere with objective user. This problem arises from increasing node deployment density in UDN. And serious interference generates from neighboring inter-cell. With respect to this problem, a novel clustering priority-based user-centric interference coordination scheme (CPUCIC) is presented, where performance gain can be attained with base station cooperation and interference nulling. Simulation results demonstrate that the proposed scheme can increase two following performance, signal interference plus noise ratio (SINR) and user bandwidth gain-to-loss ratio. However, system performance with CPUCIC scheme is increased at the cost of bandwidth.

A novel synthetic aperture radar (SAR) target recognition algorithm based upon the enhanced kernel sparse representation of monogenic signal is presented in this work. It contains two parts. First, to capture spatial and spectral properties of a target at the same time, a multi-scale monogenic feature extraction scheme is proposed. In the second module, an enhanced kernel sparse representation-based classifier (KSRC) is designed. Different from the traditional KSRC, in the enhanced KSRC, we first integrate the KPCA as well as the kernel fisher discriminant analysis (KFDA) to generate an augmented pseudo-transformation matrix. Then, a new discriminative feature mapping approach is presented by exploiting the augmented pseudo-transformation matrix so that the dimension of the kernel feature space can be effectively reduced. At last, the ℓ₁-norm minimization is utilized to calculate the sparse coefficients for a test sample, and thus the inference can be reached by the rule of minimizing total reconstruction error. Experiments on the public moving and stationary target acquisition and recognition dataset demonstrate that the proposed method achieves high recognition accuracy for SAR automatic object recognition.

FBMC can relax the out-of-band emissions problem in OFDM by using well-designed prototype filter such as the isotropic orthogonal transform algorithm. In order to achieve more flexibility in filter performances as well as to improve FBMC frequency selectivity, this paper puts forward a design method of nearly perfect reconstruction (NPR) prototype filter, in which the Nyquist condition can be constrained by using the filter autocorrelation coefficients. Then, we can apply the spectral factorization to retrieve the final prototype filter, and the filter results show advantages over conventional ones in terms of sidelobe suppression, especially in the region near the transition band. Such advantages finally result in bit-error-rate (BER) superiority in FBMC simulations, which definitely confirms the effectiveness of the proposed design method for NPR prototype filter.

Cognitive radio technologies are conceived as an emerging way to deal with the problem of current spectrum crisis. However, due to the growth of low power consumption devices and the increasing complexity of network structures in massive Internet of Things, 5G and B5G communication scenarios, energy efficiency also faces the serious challenges and attracts more attention. To enhance spectrum efficiency (SE) and energy efficiency (EE), and thus to achieve green cognitive communication in the future complex wireless networks, we consider a novel multichannel network model in which cognitive users are incorporated with the capacity of opportunistically harvesting radio energy in this paper. In this framework structure, cognitive radio users adopt the dual cooperative spectrum sensing scheme (DCS) to periodically sense the status of primary users whether exist or not in multi-bands, and harvest the radio frequency energy from primary users when they transmit data, or else, cognitive radio users can occupancy this frequency. Then, formulate spectrum and energy efficiency function with respect to transmission power, dynamic cooperative sensing time, and channel state. Using spectrum sharing and convex optimization techniques, robust optimal power and channel allocations are proposed. The experimental results show that the proposed DCS scheme with the capability of energy harvesting significantly enhance the spectrum-energy efficiency compared with another schemes.

This paper investigates the optimal precoding design for the line-of-sight (LoS) Multi-input Multi-output (MIMO) channels with the spherical-wave model (SWM). By using the SWM, we model the LoS channels with the uniform linear arrays (ULAs) at the transmitter and the uniform circular arrays (UCAs) at the receiver. Based on the channel capacity criterion, the analytical expressions for the precoding matrix and power allocation are derived. The precoding matrix and power allocation are determined by the transceiver distance, wavelength, array element spacing and the number of receive antennas. Both the analytical and simulation results show that compared with equal power allocation scheme, the proposed precoding design obtains better performance. Meanwhile, simulation results confirm the analytical results.

Capillary networks are regarded as essential compliments to cellular networks for future smart city applications. Gateways play a critical role to interconnect capillary networks with the Internet. However, traditional purpose-built hardware based gateways are not flexible enough to accommodate the emerging diverse radio access technologies and satisfy the ever-increasing networking demands. To tackle this issue, we explore the recently emerging software defined networking and network function virtualization techniques and propose a virtual gateway concept for capillary network in this paper.

Cloud radio access network (C-RAN) has been widely regarded as a promising techniques for 5G cellular mobile communication. By decoupling radio and baseband processing from all-in-one macro base station into remote radio head (RRH) and baseband unit (BBU) pool, C-RAN can significantly improve the flexibility and scalability of cellular mobile system with less ownership cost and operational expenditure. In 5G communication systems, the delay tolerance is very strict and usually limited within 5 ms. On the other hand, applications hold different delay tolerance in order to provide proper quality of service (QoS). In order to qualify the QoS requirement for different 5G applications, we classify mobile applications with different priorities, each of which holds a specified delay constraint. Based on network calculus, a theory for deterministic queuing systems, we build a traceable performance analysis framework, which reveals delay upper-bounds for applications with different priorities. With the derived upper-bounds, we can further obtain the minimum required processing capacity of the C-RAN system. Numerical analysis are conducted to validate the performance analysis framework. The experiment results show that the proposed performance analysis framework holds great potential to guide the mobile network operator for C-RAN deployment and operation.

Handling images captured in hazy weather conditions is a big challenge, for they usually have poor visibility because of the light scattering as well as absorption effects. This paper addresses an integration of a degradation model constructor and a group-based sparse representation (GSR) strategy for single image dehazing. The degradation model is constructed based on a classical physical model, i.e., dichromatic atmospheric scattering model. Then, the new degradation model is integrated into the group-based sparse representation framework. Finally, the single image dehazing problem is regarded as an image restoration problem, which can be well optimized by GSR. The method is compared with several well-known algorithms from the literature using qualitative and quantitative evaluations, and results indicate considerable improvement over existing algorithms.

Ultra-reliable low latency communications (URLLC) is one of the most important scenarios in 5G. URLLC with strict latency and reliability requirements is widely used in delay-sensitive applications such as self-driving. As the 3GPP claimed, the URLLC is amenable to 99.999% transmission correctness and within 1 ms delay bound. How to meet the requirements of reliability and latency is still an open issue. Few efforts have been made on applying a theoretical method to analyze the delay bound. Stochastic network calculus is an elegant way to obtain the delay bound based on traffic models and service guarantees. In this paper, we take the character of 5G architecture into account and use the stochastic network calculus to analyze the delay in URLLC. A tandem model is built to simulate the generation of delay in 5G network. Some factors which can influence on the delay are obtained. Numerical results are presented to verify the correctness of the delay analysis. Optimizing these factors to reduce the delay will provide valuable guidelines for the early design of URLLC architecture.