Workshops

The 1st International Workshop on Pervasive Network Intelligence for 6G Networks (PerAI-6G 2022)

Session PerAI-6G-OS

Opening Session

Conference
10:00 AM — 10:10 AM EDT
Local
May 2 Mon, 9:00 AM — 9:10 AM CDT

Session Chair

Wen Wu (Peng Cheng Laboratory, P.R. China)

Session PerAI-6G-KS1

Keynote Session 1

Conference
10:10 AM — 11:00 AM EDT
Local
May 2 Mon, 9:10 AM — 10:00 AM CDT

Data driven spectrum management for 6G Networks

Takeo Fujii (The University of Electro-Communications, Japan)

0
Finding spectrum is a key issue for realizing 6G and future wireless networks. Dynamic spectrum allocation is one of the solutions, and several systems like CBRS (citizens broadband radio service) based on SAS (spectrum access system) and AFC (automated frequency coordination) for future 6GHz wireless LAN are discussed. In the era of 6G networks, more aggressive dynamic spectrum assignment will be considered for spectrum sharing among multi-operator of 6G networks, between 6G networks and NTN (non terrestrial networks), and so on. AI technologies have attracted attention as solvers of complicated spectrum management among multiple systems considering the demand of stakeholders and the current status of the wireless environment. In this talk, data driven spectrum management for targeting 6G networks is introduced. The exact wireless environment can be estimated based on data including measurement data of personal mobile terminals and non-measurement data like 3D geographical maps including the structure of buildings. The possibility of the new style of spectrum management with AI technologies is explored to realize the sustainable development of the future wireless world.

Session Chair

Ning Zhang (University of Windsor, Canada)

Session PerAI-6G-S1

Session 1: Distributed Learning in 6G

Conference
11:00 AM — 12:00 PM EDT
Local
May 2 Mon, 10:00 AM — 11:00 AM CDT

Multi-frame Scheduling for Federated Learning over Energy-Efficient 6G Wireless Networks

Mahdi Beitollahi and Ning Lu (Queen's University, Canada)

0
It is envisioned that data-driven distributed learning approaches such as federated learning (FL) will be a key enabler for 6G wireless networks. However, the deployment of FL over wireless networks suffers from considerable energy consumption on communications and computation, which is challenging to meet stringent energy-efficiency goals of future sustainable 6G networks. In this paper, we investigate the energy consumption of transmitting scheduling decisions for FL deployed over a wireless network where mobile devices upload their local model to a coordinator (6G base station) for computing a global machine learning (ML) model iteratively. We consider that the coordinators have stringent energy efficiency goals. Therefore, to reduce the energy consumption due to the deployment of FL, we propose a novel multi-frame framework for FL that enables the coordinator to schedule wireless devices in one global round by only sending scheduling decisions at the beginning of each global round and setting the coordinator's transmission module to sleep mode to save power. In particular, we formulate a mixed-integer non-linear problem (MINLP) to minimize the average collection time of all device's local models by considering transmission errors. Then, we provide a novel method to solve the MINLP approximately and schedule wireless devices and allocate network resources. We demonstrate that our framework can save about 15 to 20 percent in some specific settings. Simulation results also show that our proposed algorithm outperforms traditional resource allocation methods and saves about 10% battery life per hundred global rounds in mobile device coordinators under certain scenarios.

Semi-Federated Learning: An Integrated Framework for Pervasive Intelligence in 6G Networks

Jingheng Zheng (Beijing University of Posts and Telecommunications, China); Wanli Ni (Beijng University of Posts and Telecommunications, China); Hui Tian (Beijng University of posts and telecommunications, China); Deniz Gündüz (Imperial College London, United Kingdom (Great Britain)); Tony Q. S. Quek (Singapore University of Technology and Design, Singapore)

1
In cellular-based federated learning (FL), the base station (BS) is only used to aggregate parameters, which incurs a waste of computing resources at the BS. In this paper, a novel semi-federated learning (SemiFL) framework is proposed to break this bottleneck, where local devices simultaneously send their gradient updates and training samples to the BS for global model computation. To capture the performance of SemiFL over wireless networks, a closed-form convergence upper bound of SemiFL is derived. Then, a non-convex problem is formulated to improve the convergence behavior of SemiFL, subject to the transmit power, communication latency, and computation distortion. To solve this intractable problem, a two-stage algorithm is proposed by controlling the transmit power and receive beamformers. Numerical experiments validate that the proposed SemiFL framework can effectively improve accuracy and accelerate convergence as compared to conventional FL.

General Decentralized Federated Learning for Communication-Computation Tradeoff

Wei Liu (University of Science and Technology of China, China); Chen Li (University of Science And Technology of China, China); Weidong Wang (University of Science and Technology of China, China)

0
Decentralized stochastic gradient descent (SGD) is a driving engine for decentralized federated learning (DFL). The performance of decentralized SGD is jointly influenced by inter-node communications and local updates. In this paper, we propose a general DFL framework, which implements both multiple local updates and multiple inter-node communications periodically, to strike a balance between communication efficiency and model consensus. It can provide a general decentralized SGD analytical framework. We establish strong convergence guarantees for the proposed DFL algorithm without the assumption of convex objectives. Based on the balance of communication and computing costs, the convergence upper bound of DFL can be depicted explicitly. Experiment results based on MNIST and CIFAR-10 datasets illustrate the superiority of DFL over traditional decentralized SGD methods.

Session Chair

Peng Yang (Huazhong University of Science and Technology, P.R. China)

Session Break Session

Break Session

Conference
12:00 PM — 12:30 PM EDT
Local
May 2 Mon, 11:00 AM — 11:30 AM CDT

Session PerAI-6G-S2

Session 2: Resource Management in 6G

Conference
12:30 PM — 2:00 PM EDT
Local
May 2 Mon, 11:30 AM — 1:00 PM CDT

DRL-based Beam Allocation in Relay-aided Multi-user MmWave Vehicular Networks

Ying Ju (Xidian University, China); Haoyu Wang (University of California Irvine, US); Yuchao Chen and Lei Liu (Xidian University, China); Tong-Xing Zheng (Xi'an Jiaotong Unviersity, China); Qingqi Pei (Xidian University, China); Ming Xiao (Royal Institute of Technology, Sweden)

1
Millimeter wave (mmWave) communication can realize high transmission rates in vehicular networks. Nevertheless, severe blocking effects and high mobility of vehicles would seriously affect downlink services for vehicles. To ensure communication quality and stability, this paper jointly explores beam allocation and relay selection in mmWave vehicular networks from the perspective of artificial intelligence-driven model. We utilize queuing theory to simulate dynamic distributions of vehicles and firstly propose a deep reinforcement learning (DRL) based joint beam allocation and relay selection scheme to mitigate the blocking effects and optimize the total communication capacity. When the expected downlink is blocked, mmWave base station (mmBS) can select appropriate idle vehicles as the relay nodes for service. Besides, we set the capacity threshold when designing the scheme to guarantee each target vehicle can obtain the ideal service. Through proper training, mmBS can intelligently find an optimal solution for the constantly updated vehicular networks based on the location of vehicles. Simulation results demonstrate the effectiveness of our scheme, which can restrain the transmission outage caused by random blockage and improve the total communication capacity of vehicular networks.

DRL-Based Fountain Codes for Concurrent Multipath Transfer in 6G Networks

Chengxiao Yu, Wei Quan, Liu Kang, Mingyuan Liu, Ziheng Xu and Hongke Zhang (Beijing Jiaotong University, China)

0
Concurrent multipath transfer (CMT) has greatly potential to significantly improve the end-to-end throughout with its multihoming property. However, due to the extremely high unpredictability of 6G heterogeneous networks, the receive buffer blocking problem seriously degrades the overall transmission reliability. To address this problem, a learning-based fountain codes scheme for concurrent multipath transfer (CMT-FC) is proposed to mitigate the negative influence of the path diversity for 6G heterogeneous networks. Specifically, we first formulate a multi-dimensional optimal problem to mitigate receive buffer blocking phenomenon and improve the transmission rate with requirement constrains. Then, we transform the data scheduling and redundancy coding rate problem into a Markov decision process, and propose a deep reinforcement learning (DRL)-based traffic management algorithm to dynamically adjust data scheduling policy and redundancy coding rate. Extensive experiments indicate the proposed algorithm mitigates the packet out-of-order problem, and improves the average throughput compared with traditional multipath transmission scheme.

Wireless Resource Management in Intelligent Semantic Communication Networks

Le Xia and Yao Sun (University of Glasgow, United Kingdom (Great Britain)); Xiaoqian Li and Gang Feng (University of Electronic Science and Technology of China, China); Muhammad Ali Imran (University of Glasgow, United Kingdom (Great Britain))

0
The prosperity of artificial intelligence (AI) has laid a promising paradigm of communication system, i.e., intelligent semantic communication (ISC), where semantic contents, instead of traditional bit sequences, are coded by AI models for efficient communication. Due to the unique demand of background knowledge for semantic recovery, wireless resource management faces new challenges in ISC. In this paper, we address the user association (UA) and bandwidth allocation (BA) problems in an ISC-enabled heterogeneous network (ISC-HetNet). We first introduce the auxiliary knowledge base (KB) into the system model, and develop a new performance metric for the ISC-HetNet, named system throughput in message (STM). Joint optimization of UA and BA is then formulated with the aim of STM maximization subject to KB matching and wireless bandwidth constraints. To this end, we propose a two-stage solution, including a stochastic programming method in the first stage to obtain a deterministic objective with semantic confidence, and a heuristic algorithm in the second stage to reach the optimality of UA and BA. Numerical results show great superiority and reliability of our proposed solution on the STM performance when compared with two baseline algorithms.

Dynamic Data Offloading for Massive Users in Ultra-dense LEO Satellite Networks based on Stackelberg Mean Field Game

Dezhi Wang and Wei Wang (Zhejiang University, China); Yuhan Kang and Zhu Han (University of Houston, USA)

1
The low earth orbit (LEO) satellite networks are seen as a promising technology to provide seamless services to remote areas, such as rural areas. In this paper, we consider the dynamic data offloading problem in ultra-dense LEO satellites networks (such as SpaceX Starlink) with large-scale users, where each user makes the offloading decision considering its state information, the influence from other users, and the prices pay to satellites. To investigate the interactive problem between users and satellites, we adopt the Stackelberg game algorithm to formulate the problem. Specifically, the satellites are leaders and decide the prices to compute the task at each time slot. On the contrary, the users are followers and decide the power allocation according to their states, the influence from other users and the prices pay to satellites. Then, since the influence is difficult to consider due to the large-scale users, we adopt the mean field game algorithm to transform the influence from others, and satellites into the mean field term, and reformulate the optimization problem as Stackelberg mean field game (SMFG) problem. Next, we transform the Fokker-Planck-Kolmogorov (FPK) equation into a linear form via Taylor expansion and adopt the G-prox primal-dual hybrid gradient (PDHG) algorithm to solve the optimization problem for users. In addition, we adopt the adjoint algorithm to solve the optimization problem for satellites. In the end, the numerical results show the effectiveness of the proposed algorithm.

Session Chair

Ruozhou Yu (North Carolina State University, United States)

Session PerAI-6G-KS2

Keynote Session 2

Conference
3:00 PM — 3:50 PM EDT
Local
May 2 Mon, 2:00 PM — 2:50 PM CDT

Digital Twin-enabled Ultra-Reliable and Low-Latency 6G for Industrial Automation

Trung Q. Duong (Queen's University Belfast (Queen's University Belfast, UK)

0
It is expected that there will be 100 Billion Internet-of-Things devices by the year 2025. Thus, the need for improved wireless reliability and latency is greater than ever. However, implementing algorithms to ensure low-latency communication for massive numbers of power-constrained mobile devices conflicts directly with the need for ultra-reliability. Recent advances in communication technologies and powerful computation platforms open opportunities to implement a wide range of breakthrough applications, especially for time sensitive services in industrial automation. In terms of communication perspective, 6G with ultra-reliable and low latency communications (URLLC) will play a vital role in the development and deployment of mission-critical applications, which require high demands on reliability and low latency communications. This opens opportunities to enable a wide range of new applications such as virtual reality (VR) with a 360-degree view, factory automation, autonomous vehicles, remote healthcare, etc. In addition, the development of digital twin opens new opportunities for transforming the cyber-physical systems in terms of intelligence, efficiency and flexibility. However, there are still many technical issues to be resolved to achieve high reliability and low latency with digital twin and apply this technology in practical scenarios due to the complexity of resource allocation in short packet transmissions. This talk will discuss digital twin technologies in industrial automation that require high data rates with ultra-reliability at very low latency for which URLLC is a natural choice.

Session Chair

Qiang Ye (Memorial University of Newfoundland, Canada)

Session PerAI-6G-S3

Session 3: AI applications in 6G

Conference
3:50 PM — 5:00 PM EDT
Local
May 2 Mon, 2:50 PM — 4:00 PM CDT

Adversarial Attacks on Deep Neural Networks Based Modulation Recognition

Mingqian Liu and Zhenju Zhang (Xidian University, China); Nan Zhao (Dalian University of Technology, China); Yunfei Chen (University of Warwick, United Kingdom (Great Britain))

0
Modulation recognition models based on deep neural network (DNN) have the advantages of automatic feature extraction, fast recognition speed and high recognition accuracy. However, due to the interpretability defects, DNN models are vulnerable to adversarial examples designed by attackers. However, most of the existing researches only focus on the accuracy of modulation recognition models, while ignoring the huge threat of adversarial examples to the safety and reliability of the models. In the field of modulation recognition, many existing attack methods have good attack performance on simple neural networks, but poor performance on excellent DNNs. Therefore, this paper proposes an adversarial attack method based on dynamic iterative. The proposed method uses a dynamic iterative step that changes with iteration instead of the fixed iterative step. Simulation results show that the proposed attack method has better attack performance when the disturbance is specified compared with the traditional attack methods.

Intelligent Sensing and Communication assisted Pedestrians Recognition in Vehicular Networks: An Effective Throughput Maximization Approach

Dengfeng Yao, Minghui Dai, Tianshun Wang and Yuan Wu (University of Macau, Macao); Zhou Su (Xi'an Jiaotong University)

0
Intelligent vehicular network has been envisioned as an important paradigm of future pervasive intelligent networks in the sixth generation (6G) systems. To improve the efficiency of sensing and communication in future intelligent vehicular networks, the integrated sensing and communication (ISAC), which combines the communication and radar modules, has recently emerged as a promising scheme to improve spectrum efficiency by sharing bandwidth for radar sensing and data communication. In this paper, we investigate the intelligent ISAC for the scenario where the recognition targets are the same as the communication targets, namely, the vehicular transmitter first uses radar sensing to detect the potential pedestrian receivers and then sends data to those detected receivers. In particular, the sensing accuracy influences the consequent effective throughput to the detected users, which thus motivates us to formulate a joint allocation scheme of sensing-slot and transmission-duration for multi-user intelligent ISAC vehicular networks, with the objective of maximizing the overall effective throughput while ensuring the fairness among the users. Despite the nature of mixed integer and non-convex programming problem, we propose a layered approach in which we firstly optimize the transmission-durations under a given sensing-slot allocation. Then, we optimize the sensing-slot allocation by proposing an myopic allocation algorithm. Finally, we provide simulation results to validate the efficiency and effectiveness of our proposed algorithm, in comparison with some benchmark schemes.

Connectivity-Aware Fast Network Forming Aided By Digital Twin For Emergency Use

Terry Guo (Tennessee Tech University, USA)

0
This paper studies a 6G use case for connecting users in emergency. A fast-forming network with a resource-limited aerial base station (BS) connecting with scattered communities of users is considered. Properly connecting each individual user in the service area is achieved by 1) applying device-to-device (D2D) communication within each community, 2) fairly assigning limited number of channels, and 3) optimally placing the flying BS. Q-learning, a common type of reinforcement learning (RL), is employed for autonomous BS placement. Two optimization objectives for BS placement are considered to maximize the per-user data rate in the worse condition and minimize the total BS transmitted power, respectively. To overcome resource limitations of the aerial BS, the RL training with many iterations is a done in the digital twin (DT) virtual space connected to the physical space via an aerial BS enabled by 6G technology. It is shown that, even the models used in DT is imperfect, nearly-optimal results can be obtained in DT. In particular, compared to RL training 100% in the physical space, a huge number of BS moves can be avoided and significant among of time and energy can be saved.

Session Chair

Yu Cheng (Illinois Institute of Technology, United States)

Session PerAI-6G-CS

Closing Session

Conference
5:00 PM — 5:10 PM EDT
Local
May 2 Mon, 4:00 PM — 4:10 PM CDT

Session Chair

Yu Cheng (Illinois Institute of Technology, United States)

Made with in Toronto · Privacy Policy · INFOCOM 2020 · INFOCOM 2021 · © 2022 Duetone Corp.