Welcome to IEEE Network - The Magazine of Global Internetworking. IEEE Network, published bimonthly, offers readers topics of interest to the networking community. As such, IEEE Network provides a focus for highlighting and discussing major computer communications issues and developments. The articles are intended to be surveys or tutorials, slanted towards the practical, and comprehensible to the nonspecialist, as well as practitioners.

Special Issue on Native Artificial Intelligence in Integrated Terrestrial and Non-Terrestrial Networks in 6G
Submission Date: 2022-05-21

The upsurge of interest in the sixth generation (6G) wireless networks, driven by the emergence of novel data-hungry applications, such as virtual/augmented/mixed reality (VR/AR/MR) services, tactile internet, haptic applications, autonomous systems, and holographic-type communications, is pushing the current infrastructure to its limits. This necessitates a radical departure from the conventional ground communications to innovative paradigms, such as integrated terrestrial and non-terrestrial networks (TNTNs), which incorporates terrestrial, air and space layers to enable multi-layer communications and extend the current network capabilities and resources. Albeit the potential advantages, in terms of throughput, coverage, and resilience, the deployment of integrated TNTNs poses new challenges, pertaining to the enormous amount of data and network traffic produced, exchanged, and managed in both inter- and intra- layer communications.

It has been shown recently that the uniqueness of 6G networks, compared to previous wireless network generations, lies in the realization of ubiquitous intelligence, in which native artificial intelligence (AI) will be the key to orchestrate wireless networks from the core to the edge, and to the cloud. To this end, machine learning (ML), which is a subfield of AI, is anticipated to be an indispensable tool in future 6G networks, which operates on the data collected from all network segments in order to enable smart resource management, access control, and multi-layer communications. However, the anticipated vision for 6G networks goes beyond leveraging ML to replace particular modules in the network. Rather, it is envisioned that each network node will enjoy a level of intelligence that enables it to continuously learn from the environment, and therefore, adapt to the network changes. The inherent heterogeneous characteristics/requirements of different nodes in each layer and among different layers noticeably exacerbate the communication management and coordination difficulty, owing to the resulted heterogeneous data.  Additionally, in conventional ML algorithms, raw data generated and stored at local  devices should be  sent  to  centralized servers for processing, training, and aggregation, yielding compromised users’ privacy and security, and increased network overhead. Furthermore, centralized ML (CML) suffers from long propagation delay, rendering it unsuitable for real-time applications. These challenges are particularly pronounced in integrated TNTNs.

Motivated by the increasing demand for secure AI tools, and the enhanced on-board computing and storage capabilities of wireless devices, the research has started to shift from centralized to distributed learning approaches. In this respect, distributed ML (DML), including the federated learning (FL), has been recently identified as an enabling technology that is capable of training wireless networks without leaking private information or consuming network resources [4]. In particular, DML allows a set of local devices to locally and collaboratively participate in the training process of a global model without having to upload their raw local data to centralized servers.

Although DML has received significant attention in the context of wireless networks, the research on the implementation of DML in single and multiple layers in integrated TNTNs is still in its infancy. In particular, several design aspects and challenges, pertaining to inter- and intra- layer communication, including but are not limited to, client selection and scheduling, joint communication and learning, model aggregation and compression, data imbalance, model convergence rate, and resource allocation, are yet to be addressed.

The objective of this Special Issue (SI) is to solicit research papers with original contributions that address the latest advances and challenges in distributed and centralized ML-enabled satellite, aerial, ground, and integrated networks, paving the way for the efficient realization and integration of DML and CML in future 6G networks. More specifically, this SI will bring together leading researchers from both industry and academia to present their views on this emerging research with respect to the fundamentals, core design aspects, applications, use-cases, and challenges of CML and DML empowered wireless networks. The papers will be peer reviewed by at least three independent experts and will be selected on their relevance to the theme of this SI.

Topics of interest include, but are not limited to:

    Architecture design and algorithms of DML & CML
    CML/DML-enabled ground communications
    CML/DML-enabled aerial networks
    CML/DML-enabled satellite networks
    CML/DML-enabled integrated TNTNs
    CML/DML for secure PHY-layer
    CML/DML-enabled TNTNs for IoT
    CML/DML-based trajectory optimization in Aerial networks
    Asynchronous CML/DML for edge devices
    Client selection and scheduling in DML
    Joint communication, sensing, and learning
    Model compression and aggregation in DML
    Efficient schemes for data imbalance in CML/DML
    CML/DML for channel modeling and estimation in integrated TNTNs
    Privacy and efficiency trade-off in CML/DML
    Enhanced security, privacy, and trust in TNTNs through CML/DML
    The interplay of CML/DML and blockchain for secure TNTNs
    Distributed and centralized machine learning algorithms in mobile edge computing
    Distributed and centralized machine learning empirical studies
    Distributed and centralized machine learning applications in TNTNs
    Energy efficient techniques for improved TNTNs through CML/DML
    CML/DML in emerging applications
    Network protocol designs for CML/DML in terrestrial, aerial, and satellite networks
    Optimization of CML/DML-enabled TNTNs
    Efficient schemes for data heterogeneity and dependency in DML-enabled networks
    Meta-learning
    Multi-task optimization and learning
    CML/DML for resource management in TNTNs
    The interplay between CML/DML and reconfigurable intelligent surfaces in TNTNs
    Federated learning-enabled TNTNs
    Incentive mechanisms design and game-theoretic approaches for ML-enabled TNTNs
    CML/DML and optical wireless communication in satellite, aerial, and terrestrial networks.

The topics covered by the proposed IEEE Network SI are aimed to be the foundation for the revolution of new CML & DML paradigms to be implemented vertically over multiple layers, including, the space, air, and ground.

Submission Guidelines

Manuscripts should conform to the standard format as indicated in the "Information for Authors" section of the Paper Submission Guidelines.

All manuscripts to be considered for publication must be submitted by the deadline through Manuscript Central. Select “November 2022/AI-TNTN” from the drop-down menu of topic titles.

Important Dates

Manuscript Submission Deadline: 31 May 2022
Initial Decision Notification: 15 July 2022
Revised Manuscript Due: 15 August 2022
Final Decision Notification: 31 August 2022
Final Manuscript Due: 15 September 2022
Publication Date: November/December 2022

Guest Editors

Lina Bariah
KU C2PS, Khalifa University
UAE/University at Albany, SUNY, USA

Sami Muhaidat
KU C2PS, Khalifa University
UAE/Carleton University, Canada

Daniel Benevides Da Costa
Technology Innovation Institute, UAE

Guosen Yue
Futurewei Technologies, Inc, USA

Ekram Hossain
University of Manitoba, Canada

Merouane Debbah
Technology Innovation Institute, UAE, and
Lagrange Mathematical and Computing Research Center, France