Ubiquitous sensors and smart devices from factories and communities are generating massive amounts of data, and ever-increasing computing power is driving the core of computation and services from the cloud to the edge of the network. As an important enabler broadly changing people's lives, from face recognition to ambitious smart factories and cities, developments of artificial intelligence (especially deep learning, DL) based applications and services are thriving. However, due to efficiency and latency issues, the current cloud computing service architecture hinders the vision of "providing artificial intelligence for every person and every organization at everywhere". Thus, unleashing DL services using resources at the network edge near the data sources has emerged as a desirable solution. Therefore, edge intelligence, aiming to facilitate the deployment of DL services by edge computing, has received significant attention. In addition, DL, as the representative technique of artificial intelligence, can be integrated into edge computing frameworks to build intelligent edge for dynamic, adaptive edge maintenance and management. With regard to mutually beneficial edge intelligence and intelligent edge, this paper introduces and discusses: 1) the application scenarios of both; 2) the practical implementation methods and enabling technologies, namely DL training and inference in the customized edge computing framework; 3) challenges and future trends of more pervasive and fine-grained intelligence. We believe that by consolidating information scattered across the communication, networking, and DL areas, this survey can help readers to understand the connections between enabling technologies while promoting further discussions on the fusion of edge intelligence and intelligent edge, i.e., Edge DL.

Convergence of Edge Computing and Deep Learning: A Comprehensive Survey

Cooperation between the fog and the cloud in mobile cloud computing environments could offer improved offloading services to smart mobile user equipment (UE) with computation intensive tasks. In this paper, we tackle the computation offloading problem in a mixed fog/cloud system by jointly optimizing the offloading decisions and the allocation of computation resource, transmit power, and radio bandwidth while guaranteeing user fairness and maximum tolerable delay. This optimization problem is formulated to minimize the maximal weighted cost of delay and energy consumption (EC) among all UEs, which is a mixed-integer non-linear programming problem. Due to the NP-hardness of the problem, we propose a low-complexity suboptimal algorithm to solve it, where the offloading decisions are obtained via semidefinite relaxation and randomization, and the resource allocation is obtained using fractional programming theory and Lagrangian dual decomposition. Simulation results are presented to verify the convergence performance of our proposed algorithms and their achieved fairness among UEs, and the performance gains in terms of delay, EC, and the number of beneficial UEs over existing algorithms.

/pdf/computation-offloading-and-resource-allocation-in-mixed-fog-3ht8oi2zwm.pdf

Computation Offloading and Resource Allocation in Mixed Fog/Cloud Computing Systems With Min-Max Fairness Guarantee

Sending data to the cloud for analysis was a prominent trend during the past decades, driving cloud computing as a dominant computing paradigm. However, the dramatically increasing number of devices and data traffic in the Internet-of-Things (IoT) era are posing significant burdens on the capacity-limited Internet and uncontrollable service delay. It becomes difficult to meet the delay-sensitive and context-aware service requirements of IoT applications by using cloud computing alone. Facing these challenges, computing paradigms are shifting from the centralized cloud computing to distributed edge computing. Several new computing paradigms, including Transparent Computing, Mobile Edge Computing, Fog Computing, and Cloudlet, have emerged to leverage the distributed resources at network edge to provide timely and context-aware services. By integrating end devices, edge servers, and cloud, they form a hierarchical IoT architecture, i.e., End-Edge-Cloud orchestrated architecture to improve the performance of IoT systems. This article presents a comprehensive survey of these emerging computing paradigms from the perspective of end-edge-cloud orchestration. Specifically, we first introduce and compare the architectures and characteristics of different computing paradigms. Then, a comprehensive survey is presented to discuss state-of-the-art research in terms of computation offloading, caching, security, and privacy. Finally, some potential research directions are envisioned for fostering continuous research efforts.

A Survey on End-Edge-Cloud Orchestrated Network Computing Paradigms: Transparent Computing, Mobile Edge Computing, Fog Computing, and Cloudlet

Edge caching is an emerging technology for addressing massive content access in mobile networks to support rapidly growing Internet-of-Things (IoT) services and applications. However, most current optimization-based methods lack a self-adaptive ability in dynamic environments. To tackle these challenges, current learning-based approaches are generally proposed in a centralized way. However, network resources may be overconsumed during the training and data transmission process. To address the complex and dynamic control issues, we propose a federated deep-reinforcement-learning-based cooperative edge caching (FADE) framework. FADE enables base stations (BSs) to cooperatively learn a shared predictive model by considering the first-round training parameters of the BSs as the initial input of the local training, and then uploads near-optimal local parameters to the BSs to participate in the next round of global training. Furthermore, we prove the expectation convergence of FADE. Trace-driven simulation results demonstrate the effectiveness of the proposed FADE framework on reducing the performance loss and average delay, offloading backhaul traffic, and improving the hit rate.

/pdf/federated-deep-reinforcement-learning-for-internet-of-things-2rmvlrlj8p.pdf

Federated Deep Reinforcement Learning for Internet of Things With Decentralized Cooperative Edge Caching

Mobile data traffic is currently growing exponentially and these rapid increases have caused the backhaul data rate requirements to become the major bottleneck to reducing costs and raising revenue for operators. To address this problem, caching techniques have attracted significant attention since they can effectively reduce the backhaul traffic by eliminating duplicate data transmission that carries popular content. In addition, other system performance metrics can also be improved through caching techniques, e.g., spectrum efficiency, energy efficiency, and transmission delay. In this paper, we provide a systematical survey of the state-of-the-art caching techniques that were recently developed in cellular networks, including macro-cellular networks, heterogeneous networks, device-to-device networks, cloud-radio access networks, and fog-radio access networks. In particular, we give a tutorial on the fundamental caching techniques and introduce caching algorithms from three aspects, i.e., content placement, content delivery, and joint placement and delivery. We provide comprehensive comparisons among different algorithms in terms of different performance metrics, including throughput, backhaul cost, power consumption, and network delay. Finally, we summarize the main research achievements in different networks, and highlight main challenges and potential research directions.

https://ieeexplore.ieee.org/ielaam/9739/8443345/8327582-aam.pdf

A Survey of Caching Techniques in Cellular Networks: Research Issues and Challenges in Content Placement and Delivery Strategies

Content caching at base stations (BSs) is a promising technique for future wireless networks by reducing network traffic and alleviating server bottleneck. However, in practice, the content popularity distribution may change with spatio-temporal variation but be unknown for BSs, which is an intractable obstacle for efficient caching strategy design. In this paper, considering unknown popularity distribution, we explore the content caching problem by jointly optimizing the content caching in cooperative BSs, content sharing among BSs, and cost of content retrieving. We tackle the problem from a multi-armed bandit learning perspective, where the learning of the popularity distribution is incorporated with the content caching and sharing process. Specifically, we first propose a centralized algorithm by employing a semidefinite relaxation approach, and we prove that this centralized algorithm learns efficient caching by deriving a sub-linear learning regret bound. To further reduce computational complexity, we propose a distributed algorithm based on alternating direction method of multipliers, where each BS only solves their own problems by exchanging local information with neighbor BSs. Extensive simulation results show the effectiveness of the proposed algorithms in terms of learning content popularity distributions of individual BSs, offloading traffic from the content server, and reducing cost of content retrieving.

Learning-Based Content Caching and Sharing for Wireless Networks

To cater for the exploding growth of video traffic, small cell base stations (SBSs) and deviceto- device-enabled caching and delivery have been regarded as promising techniques for future wireless networks. In this article, we design a proximity communications enhanced multilayer caching and delivery architecture. Then merits possessed by the proposed architecture are highlighted, and challenges and open issues are comprehensively presented. Specifically, we shed light on the trade-offs between key performance indicators (e.g., hit ratio, latency, and coverage) and operation costs (e.g., device storage space, wireless bandwidth, and device battery life), and then clarify fundamental coupling between content caching and delivery. To further verify the effectiveness of the cooperation among SBSs and user equipments, we propose a distributed content caching and delivery strategy, jointly considering popularity distribution, diverse storage capability, and user mobility. Simulation results demonstrate that the proposed strategy can significantly lower the content retrieval latency and reduce the traffic flowing to core networks. Furthermore, design details of the experimental testbed are presented, and the validity of our developed strategy is verified.

Enhancement for content delivery with proximity communications in caching enabled wireless networks: architecture and challenges

In this letter, we investigate the fundamental tradeoff between rate and energy for sparse code multiple access (SCMA) networks with wireless power transfer. A weighted rate and energy maximization problem by jointly considering power allocation, codebook assignment, and power splitting, is formulated. To solve the hard problem, an iterative algorithm based on the univariate search technique is proposed, which has good performance with low complexity. Specifically, we analyze the special structure of the problem and exploit it to obtain the optimal power splitting ratio and resource allocation strategy when one of them is fixed. Simulation results indicate that our algorithm achieves a better rate-energy tradeoff compared to other schemes.

Rate and Energy Maximization in SCMA Networks With Wireless Information and Power Transfer

In wireless networks, the technique of caching contents cooperatively in user equipments (UEs) and base stations (BSs) enables to improve the performance of data access. In this paper, we consider cluster based cooperative device-to-device (D2D) caching enabled network and formulate the caching problem to minimize content retrieving delay. To achieve this, we develop a content caching and replacement scheme by considering the limited storage capacity of UEs, content popularity, and content access process in the network. As a result, our scheme has shown the minimal delay by caching popular contents on UEs and BS with delivering requested contents at smaller average request hops. Moreover, we evaluate our approach with other caching schemes based on parameters as the number of UEs joining the cluster, content popularity and the number of edge UE in a cluster. Simulation results show that the proposed caching scheme significantly outperforms the other schemes.

Minimum delay guaranteed cooperative device-to-device caching in 5G wireless networks

In this paper, by jointly considering subchannels, power, and Modulation and Coding Scheme (MCS) allocation, we address the sum-rate maximization problem in OFDMA downlink systems. We formulate the problem as an integer linear programming (ILP), which maximizes the system sum-rate subject to the minimum rate requirements of users and total transmit power constraint of base station. To solve the formulation with low complexity, we propose a two-level iterative Subchannels, Power, and MCS allocation Algorithm (SPMA) by exploiting Tabu Search (TS). At each iteration, the SPMA firstly assigns MCS to users and then allocates subchannels and power based on a SubChannels and Power allocation Algorithm (SCPA). Particularly, the SCPA maximizes the system sum-rate by first satisfying the minimum rate requirements with the least transmit power. Simulation results show that the SPMA outperforms the existing algorithms in terms of sum-rate and average rate per user, as well as demonstrate that the sum-rate is distributed flexibly among users in instantaneous channel conditions with the SPMA.

Jiongjiong Song

Papers

Learning-Based Content Caching and Sharing for Wireless Networks

Enhancement for content delivery with proximity communications in caching enabled wireless networks: architecture and challenges

Rate and Energy Maximization in SCMA Networks With Wireless Information and Power Transfer

Minimum delay guaranteed cooperative device-to-device caching in 5G wireless networks

Sum-rate maximization in OFDMA downlink systems: A joint subchannels, power, and MCS allocation approach