Journal of Communications and Information Networks
Print ISSN 2096-1081 CN 10-1375/TN
Editor-in-Chief: Wei Zhang
With the rapid upsurge of deep learning tasks at the network edge, effective edge artificial intelligence(AI)inference becomes critical to provide lowlatency intelligent services for mobile users via leveraging the edge computing capability. In such scenarios, energy efficiency becomes a primary concern. In this paper, we present a joint inference task selection and downlink beamforming strategy to achieve energy-efficient edge AI inference through minimizing the overall power consumption consisting of both computation and transmission power consumption, yielding a mixed combinatorial optimization problem. By exploiting the inherent connections between the set of task selection and group sparsity structural transmit beamforming vector, we reformulate the optimization as a group sparse beamforming problem. To solve this challenging problem, we propose a logsum function based three-stage approach. By adopting the log-sum function to enhance the group sparsity, a proximal iteratively reweighted algorithm is developed. Furthermore, we establish the global convergence analysis and provide the ergodic worst-case convergence rate for this algorithm. Simulation results will demonstrate the effectiveness of the proposed approach for improving energy efficiency in edge AI inference systems.
面向低功耗边缘人工智能模型推理的稀疏优化方法
随着网络边缘深度学习应用的迅速兴起，通过融合边缘计算和人工智能技术，边缘人工智能模型推理成为支持移动用户低延迟智能服务的关键技术。在这种情况下，能量效率成为首要考虑的问题。通过边缘服务器节点的协作计算和通信，本文提出了一种联合推理任务选择和下行波束赋形设计的策略，通过最小化计算功耗和传输功耗构成的总功耗，实现节能的边缘人工智能模型推理。这在数学上是一个混合组合优化问题。通过挖掘计算任务选择集合与波束赋形系数的组稀疏结构，将该优化问题转化为组稀疏波束赋形优化问题。为了解决这一具有挑战性的问题，我们提出了一种基于对数求和函数的三阶段策略。采用对数求和函数增强组稀疏性，提出了一种近似迭代加权算法。此外，我们还提供了该算法的全局收敛性分析，并给出了遍历最坏情况下的收敛速度。仿真结果将证明所提出的方法在提高边缘人工智能推理系统的能量效率方面是有效的。
As a widely-used machine-learning classifier, a decision tree model can be trained and deployed at a service provider to provide classification services for clients, e. g. , remote diagnostics. To address privacy concerns regarding the sensitive information in these services(i. e. , the clients’ inputs, model parameters, and classification results), we propose a privacy-preserving decision tree classification scheme (PDTC) in this paper. Specifically, we first tailor an additively homomorphic encryption primitive and a secret sharing technique to design a new secure two-party comparison protocol, where the numeric inputs of each party can be privately compared as a whole instead of doing that in a bit-by-bit manner. Then, based on the comparison protocol, we exploit the structure of the decision tree to construct PDTC, where the input of a client and the model parameters of a service provider are concealed from the counterparty and the classification result is only revealed to the client. A formal simulation-based security model and the security proof demonstrate that PDTC achieves desirable security properties. In addition, performance evaluation shows that PDTC achieves a lower communication and computation overhead compared with existing schemes.
安全高效的隐私保护的决策树分类
决策树作为一种被广泛应用的机器学习分类器，可以被服务提供者部署到远程服务器中并为客户提供服务，例如远程医疗诊断服务。为了解决客户及服务提供者的敏感信息（客户的输入及分类结果，服务提供者的模型参数）可能会被泄露这一隐私问题，我们提出了一个隐私保护的决策树分类方案。具体来说，我们首先利用加法同态加密原语和秘密共享技术设计了一个新的两方安全比较协议，协议中两方的数值输入可以被作为一个整体安全地进行比较，而不用以逐比特方式进行。基于提出的比较协议，我们利用决策树的树形结构构建了隐私保护的决策树分类方案（PDTC），其中客户的输入和服务提供者的模型参数都不会泄露给对方，并且决策树分类结果只会被客户知晓。基于模拟的安全模型和安全性证明论证了PDTC实现了预期的安全特性。此外，性能评估部分展示了相较于现有的方案，PDTC具有较低的通信开销和计算开销。
In order to meet the real-time performance requirements, intelligent decisions in Internet of things applications must take place right here right now at the network edge. Pushing the artificial intelligence frontier to achieve edge intelligence is nontrivial due to the constrained computing resources and limited training data at the network edge. To tackle these challenges, we develop a distributionally robust optimization(DRO)-based edge learning algorithm, where the uncertainty model is constructed to foster the synergy of cloud knowledge and local training. Specifically, the cloud transferred knowledge is in the form of a Dirichlet process prior distribution for the edge model parameters, and the edge device further constructs an uncertainty set centered around the empirical distribution of its local samples. The edge learning DRO problem, subject to these two distributional uncertainty constraints, is recast as a single-layer optimization problem using a duality approach. We then use an Expectation-Maximization algorithm-inspired method to derive a convex relaxation, based on which we devise algorithms to learn the edge model. Furthermore, we illustrate that the meta-learning fast adaptation procedure is equivalent to our proposed Dirichlet process prior-based approach. Finally, extensive experiments are implemented to showcase the performance gain over standard approaches using edge data only.
边缘网络中基于狄利克雷过程（Dirichlet Process）先验的分布鲁棒性学习
为了满足实时性能要求，物联网应用中的智能决策过程必须立即在网络边缘端进行。由于网络边缘端受限的计算资源和有限的训练数据，实现边缘智能（Edge Intelligence）并非易事。为了解决这些挑战，我们开发了一种基于分布鲁棒性优化（DRO）的边缘学习算法，其中我们构建了不确定性模型来促进云端知识与本地训练的协同作用。具体而言，云端传输的知识首先被构造成边缘端模型参数的狄利克雷过程（Dirichlet Process）先验分布的形式，边缘设备再进一步构造一个以其本地样本的经验分布为中心的不确定性集。在这两个分布的不确定性约束下，我们使用对偶法将边缘学习的DRO问题转化成单层优化问题。然后我们使用一个受最大期望算法（EM Algorithm）启发的方法来导出上述单层优化问题的凸松弛，并在此基础上设计了边缘端模型的相应具体学习算法。此外，我们还说明了元学习（Meta Learning）中的快速自适应过程等效于我们提出的基于狄利克雷过程先验的方法。最后，我们进行了大量的实验，以展示相较于仅使用边缘端数据的传统方法获得的性能提升。
As a major component of thefifth-generation (5G)wireless networks, network densification greatly increases the network capacity by adding more cell sites into the network. However, the densified network increases the handover frequency of fast-moving mobile users, like vehicles. Thus, seamless handover with security provision is highly desirable in 5G networks. The third generation partnership project (3GPP) has been working on standardization of the handover procedure in 5G networks to meet the stringent efficiency and security requirement. However, the existing handover authentication process in 5G networks has securityflaws, i. e. vulnerable to replay and de-synchronization attacks, and cannot provide perfect forward secrecy. In this paper, we propose a secure and efficient handover authentication and key management protocol utilizing the Chinese remainder theory. The proposed scheme preserves the majority part of the original 5G system architecture defined by 3GPP, thus can be easily implemented in practice. Formal security analysis based on BAN-logic shows that the proposed scheme achieves secure mutual authentication and can remedy some security flaws in original 5G handover process. Performance analysis shows that the proposed protocol has lower communication overhead and computation overhead compared with other handover authentication schemes.
一安全高效的5G网络切换认证和密钥管理协议
作为第五代(5G)移动通信网络的主要组成部分，网络密集化通过向网络中添加更多的微基站来大大增加网络容量。然而，密集化的网络增加了快速移动用户(如车辆)的网络切换频率。因此，在5G网络中，具有安全保障的无缝切换是被强烈需求的。第三代伙伴计划(3GPP)一直致力于5G网络中交接程序的标准化，以满足严格的效率和安全要求。然而现有的5G网络切换认证过程存在安全漏洞，容易受到重放和去同步攻击，且不能提供完美前向保密。本文利用中国剩余定理，提出了一种安全高效的切换认证和密钥管理协议。该方案保留了3GPP定义的5G系统的大部分原有构架，因此更易于在实践中实现。基于BAN-逻辑的形式化安全分析表明，该方案实现了安全的相互认证，并弥补了原有5G切换过程中的一些安全缺陷。性能分析表明，该方案与其他切换认证方案相比具有较低的通信开销和计算开销。
The use of unmanned aircraft systems(UAS) for critical public service missions by public organizations, local governments, and social users is expanding. Robust radio transmission by unmanned aerial vehicles (UAVs)is indispensable in these applications. One of the standards suitable for UAV wireless communications is local 5G. However, to introduce local 5G into a UAS, several problems specific to UASes must be solved. Challenges include the deterioration of the signal-to-noise ratio due to topography and obstacles, the uncertainty of the communication environment due to the movement of multiple UAVs, and the limitation of the frequency bandwidth. To address these problems and to introduce local 5G into the UAS, we propose a novel communication scheduling method. The method incorporates transmission modes with diversity. Our evaluations shows that total data volumes increase by 1. 9～3 times with the proposed scheduling method compared with transmission without diversity or efficient scheduling.
使用本地5G无人机系统的具有分集的通信调度
公共组织、地方政府和社会用户正在将无人机系统（UAS）用于重要的公共服务任务。在这些应用中，无人机（UAV）进行可靠的无线电传输是必不可少的。适用于无人机无线通信的标准之一是本地5G。但是，要将本地5G引入到UAS中，必须解决一些特定于UAS的问题。这些挑战包括由于地形和障碍物导致的信噪比下降，由于多个UAV的移动而引起的通信环境的不确定性，以及频率带宽的限制。为了解决这些问题并将本地5G引入UAS，我们提出了一种新颖的通信调度方法。该方法结合了具有分集的传输模式。我们的评估表明，与没有分集或高效调度的传输相比，使用所提出的调度方法，总数据量增加了1.9~3倍。
In this paper, we investigate the distributed antenna systems (DAS) based on device to device (DASD2D)communications under the imperfect channel state information(CSI). Our aim is to maximize the energy efficiency(EE)of the D2D users equipment(DUE)under the constraints of the maximum transmission power of D2D pairs and the quality of service(QoS)requirements of the cellular user equipment(CUE). The worst-case design is considered so that the QoS of the CUE can be guaranteed for every realization of the CSI error in the ellipsoid region. The EE objective function of the optimization problem is non-convex and non-linear, and thus this problem cannot be solved by the traditional optimization methods. To solve this problem, first we transform it to an EE maximization problem without uncertain parameters by exploiting the Markov and Cauchy-Schwartz inequality. Then using the fractional programming theory and difference of convex functions optimization method, the robust EE maximization algorithms based on the hard and soft protection method are developed to maximize the system’s EE performance, respectively. However, these two algorithms are designed at the cost of the reduced EE of the DUE. Therefore, in order to further improve the EE performance and make a trade-off between the EE performance and the robustness, the iterative update algorithms for the total power constraint and average interference constraint are developed to maximize the system’s EE performance, respectively. Simulation results demonstrate the effectiveness of the four proposed EE algorithms and illustrate the trade-off between the EE performance and robustness for the iterative update algorithms.
非完美CSI下分布式天线D2D通信系统的能量效率研究
在本文中，我们研究了在非完美信道状态信息（CSI，Channel State Information）下基于终端直通（D2D，Device to Device）通信的分布式天线系统（DAS，Distributed Antenna Systems）。我们的目标是最大化D2D用户设备（DUE，D2D Users Equipment）的能量效率（EE，Energy Efficiency），以D2D对的最大传输功率和蜂窝用户设备（CUE，Cellular User Equipment）的通信服务质量（QoS，Quality of Service）要求为约束条件。同时考虑了CSI最坏情况下的鲁棒性设计，以确保CSI落在椭球域中的任意情况都能保证CUE的QoS。由于优化问题的目标函数是非凸和非线性的，因此该问题无法通过传统的优化方法来解决。为了解决这个问题，首先我们通过利用马尔可夫（Markov）和柯西-施瓦茨（Cauchy-Schwartz）不等式将其转换为无不确定性参数的EE最大化问题。然后基于分数规划理论和凸函数差分（D.C.，Difference of Convex Functions）优化方法，分别研究了基于硬保护和软保护方法的鲁棒性EE最大化算法，以最大化系统的EE性能。但是，这两种算法是以降低DUE的EE为代价的。因此，为了进一步提高EE性能，并在EE性能和鲁棒性之间进行权衡，提出了使EE性能最大化的总功率约束和平均干扰约束的迭代更新算法。仿真结果证明了所提出的4种鲁棒性EE最大化算法的有效性，并说明了在迭代更新算法中可以在EE性能与鲁棒性之间进行权衡。
In this paper, the cooperative jobs dispatching problem in an edge computing network with multiple access points (APs) and edge servers is considered. Due to the uncertain traffic in the network between APs and edge servers, the job uploading delay can not be predicted accurately. Specifically, the job arrivals at the APs, the job uploading delay from APs to edge servers and the job computation time at the edge servers are all modeled as random variables. Since each job dispatching decision will affect the system state in the future, we formulate the joint optimization of jobs dispatching at all the APs and all the scheduling time slots as an infinite-horizon Markov decision process (MDP). The minimization objective is a discounted measurement of the average processing time per job, including the uploading delay, the waiting time and the computation time at the edge servers. In this problem, the approximate MDP should be adopted to address the curse of dimensionality. Conventional low-complexity approximate solution of MDP is usually hard to predict the performance analytically. In this paper, a novel approximate MDP solution framework is proposed via one-step policy iteration over a baseline policy, where the analytical performance bound can be obtained. Moreover, since the expression of the approximate value function is derived, the value iteration in conventional methods can be eliminated, which can essentially reduce the computation complexity. It is shown by simulations that the proposed low-complexity algorithm has significantly better performance than various benchmark schemes.
在上传时延无法预测的边缘计算网络中的协同计算任务调度
本文考虑了具有多个访问接入点（AP）和边缘服务器的边缘计算网络中协同计算任务分配问题。由于AP和边缘服务器之间的网络流量是不确定的，因此无法准确预测计算任务的上传时延。具体而言，计算任务到达AP，从AP到边缘服务器的计算任务上传时延以及边缘服务器上的任务计算时间都被建模为随机过程。由于未来每个计算任务调度决策都会影响系统状态，因此我们将所有时刻所有AP上的计算任务联合调度问题公式化为无限阶段的马尔可夫决策过程（MDP）。优化目标是最小化每个计算任务的加权平均处理时间，包括上传时延，等待时延和边缘服务器上的计算时间。在此问题中，应采用近似MDP的方法来解决“维度爆炸”问题。现有的MDP低复杂度近似解决方案通常很难分析近似算法的性能。通过对基准策略进行单步策略迭代，本文提出了一种新颖的近似MDP解决方案。该方案可以分析近似算法的性能。此外，由于推导了近似值函数的解析表达式，传统MDP中的值迭代过程可以省略，从而实质上降低了计算复杂度。仿真结果表明，本文提出的低复杂度算法比各种基准方案具有更好的性能。
Unmanned aerial vehicle (UAV)-enabled communication is a promising technology to extend coverage and enhance throughput for traditional terres-trial wireless communication systems. In this paper, we consider a UAV-enabled wireless sensor network, where a multi-antenna UAV is dispatched to collect data from a group of sensor nodes(SNs). The objective is to maximize the minimum data collection rate from all SNs via jointly optimizing their transmission scheduling and power allocations as well as the trajectory of the UAV, subject to the practical constraints on the maximum transmit power of the SNs and the maximum speed of the UAV. The formulated optimization problem is challenging to solve as it involves non-convex constraints and discrete-value variables. To draw useful insight, we first consider the special case of the formulated problem by ignoring the UAV speed constraint and optimally solve it based on the Lagrange duality method. It is shown that for this relaxed problem, the UAV should hover above a finite number of optimal locations with different durations in general. Next, we address the general case of the formulated problem where the UAV speed constraint is considered and propose a traveling salesman problem-based trajec-tory initialization, where the UAV sequentially visits the locations obtained in the relaxed problem with minimumflying time. Given this initial trajectory, we thenfind the corresponding transmission scheduling and power alloca-tions of the SNs and further optimize the UAV trajectory by applying the block coordinate descent and successive convex approximation techniques. Finally, numerical results are provided to illustrate the spectrum and energy efficiency gains of the proposed scheme for multi-antenna UAV data harvesting, as compared to benchmark schemes.
基于多天线无人机的数据收集：路径和通信资源联合优化
相较于传统的地面无线通信，无人机（UAV）辅助通信能够有效地扩大覆盖范围和提高通信吞吐量。此文研究了一种利用多天线无人机进行数据收集的无人机辅助无线传感网络。在传感器最大传输功率和无人机最大飞行速度的约束条件下，通过对传感器的传输调度、传输功率以及无人机的路径进行联合优化以最大化最小的数据收集速率。由于所建模问题涉及非凸的约束条件和离散变量，难以直接解决。为此，我们首先忽略速度的约束条件，利用拉格朗日对偶方法找到了所得松弛问题的最优解。结果显示在无速度约束的情况下，无人机只需要盘旋在数个特定的位置上方。接下来，考虑包含速度约束的原优化问题。首先引入一个基于旅行商问题的初始化路径让无人机在最短的时间内经过松弛问题中获得的位置，然后基于初始化路径，利用块坐标下降法和连续凸逼近算法来获得传感器的传输调度和传输功率，并进一步优化无人机的路径。最后，通过仿真结果验证了基于多天线无人机的数据收集在频谱效率和能量效率方面的优越性能。
Spectrum and energy resources are very important for the rapidly developing of thefifth generation (5G) wireless communication networks. Cognitive radio and energy harvesting technologies, which focus on the spectrum efficiency and the energy efficiency, respectively, can be jointly used for solving the scarcities of spectrum and energy in energy harvesting cognitive radio networks (EHCRNs), where the energy is absorbed from ambient space and spectrum is licensed to the primary users. However, how to guarantee the quality of service (QoS) for EHCRNs is still a challenging problem. In this paper, we develop the optimal power and rate adaptation scheme under statistical QoS provisioning for EHCRNs. In particular, we analyze the power constraints of EHCRNs. Then, we formulate the effective capacity maximization problem for EHCRNs, solving which we obtain the closedform of the optimal power allocation and rate adaptation scheme under different power constraints. Numerical analyses verify the effective capacity enhancement and validate the relationships among the effective capacity, the QoS exponent, the energy arrival rate, and the average interference power.
面向统计性服务质量保障的能量收集认知无线电网络资源管理
频谱资源以及能量资源在第五代无线通信网络中尤为重要，以高频谱效率为导向的认知无线电技术以及追求高能量效率的能量收集技术可用于解决能量收集认知无线电网络中频谱和能量的紧缺问题。然而，如何保障能量收集认知无线电网络中用户的服务质量（QoS）需求仍是一项有挑战性的议题。本文提出了一套面向服务质量保障的功率和速率自适应方案，可有效解决能量收集认知无线电网络中用户的服务质量保障问题。具体来说，本文分析了能量收集认知无线电网络中的功率约束来源，研究了多功率约束下的有效容量最大化问题，通过求解该问题，得到了若干不同功率约束下最优功率和速率自适应方案。仿真结果表明，提出的功率和速率自适应方案可在保障用户服务质量的前提下最大化用户的有效容量。
