The Next Horizon: Intelligent Hands-Off 6G Networks
This is the seventh blog in our 6G White Paper series looking at how technology will continue to evolve as the world adopts 6G networks.
Check out the other posts in the series here: 6G: The Next Horizon.
Imagine a communications network that essentially runs itself, that adapts and optimizes itself to achieve the best performance without the need for human intervention. This is what artificial intelligence (AI) will bring to the 6th generation of mobile networks — 6G.
AI seems to be everywhere nowadays, from smart toothbrushes to self-driving cars. While the necessity to integrate AI into some applications might not be immediately apparent, doing so in the 6G network will undoubtedly bring major benefits.
AI support in 6G
IDC estimates there will be over 55 billion IoT devices by 2025. As the number of such devices explodes and wireless sensing makes it possible to feed big data into machine learning (ML) algorithms, AI will become an engine for all types of automation.
One of the primary objectives for 6G is to support AI everywhere. 6G will run AI both as a service and as a native feature. Specifically, 6G air interface and network designs will leverage end-to-end (E2E) AI and machine learning (ML) to implement customized optimization and automated operations, administration, and management (OAM). This is known as “AI for Network (AI4NET)”.
In addition, each 6G network element will natively integrate communication, computing, and sensing capabilities, facilitating the evolution from centralized intelligence in the cloud to ubiquitous intelligence at network edges. This is known as “Network for AI (NET4AI)” or “AI as a Service (AIaaS)”.
To support AI everywhere, 6G will need to address three key challenges:
- 6G needs to be the most efficient platform for AI, using minimal capacity resources to transfer the massive quantities of big data for AI training, so that communication costs can be kept to a minimum. It will also need to implement optimally distributed computing in the network, so that we can leverage mobile edge computing while keeping computation costs low.
- 6G needs to enable the collection of massive data from the physical world (millions of times more data than at present), so that we can train ML algorithms and create a cyber world. Compressing this data to minimize the burden on the 6G network will therefore become a hot topic of research.
- 6G needs to efficiently implement distributed collaborative learning so that we can minimize the computational load involved in large-scale AI training. Data split and model split for AI will need to be integrated, and both distributed and federated learning will be required to help optimize computing resources, local learning, and global learning, while also meeting new local governance requirements for data. Core network functions will therefore migrate toward a deep-edge network, cloud-based software operations will shift toward massive ML, and the access network will shift from downlink-centric to uplink-centric.
Key benefits of AI
We know what the challenges are in bringing AI to 6G, but are they worth overcoming — will doing so be beneficial? Put simply: Yes. AI will not only enable automation, but also facilitate data management as well as distributed learning and inference.
AI-enhanced network automation
Today’s OAM of mobile networks requires a large workforce, saddling carriers with a major labor and financial burden. AI has the potential to eliminate this burden. For instance, it could enable the 6G network system to implement, operate, and manage network configurations and functions. And by using predictive network analytics and E2E system OAM across all technical domains, the 6G system would enable zero-touch proactive OAM. AI in 6G will adapt to environmental changes and optimize both the communication and computing resources for optimal solutions that meet diversified requirements.
AIaaS for data management
Huge volumes of data will be generated, collected, and exchanged in 6G. This data will be used to perform and optimize various network services related to operation and management tasks (e.g., for configuration management, fault management, and SLA assurance). The knowledge extracted from raw data could be exchanged with other systems or business sectors to unlock additional value. But while data is essential for AI, not all raw data is high quality or usable. This is where AIaaS would come into play, as it could be used to select high-quality data from vast amounts of raw data, supporting efficient data processing while also reducing computational complexity and energy consumption.
AIaaS for distributed learning and inference
As the requirements for real-time and large-scale learning and inference continue to grow into the future, AIaaS for distributed learning and inference applications will be critical for meeting these requirements.
We will see the rise of Software 2.0, where software is no longer written by humans but instead developed by AI. Here, massive data will be provided to deep learning algorithms to generate deep neural network (DNN) models for application development.
The 6G network will not simply be a big pipe that transmits bits and bytes. Instead, it will be a consolidated platform — with integrated connectivity and computing capabilities — that provides optimal resource scheduling necessary to support learning tasks and achieve fast learning convergence. The benefits of this will go beyond just the superior performance (e.g., ultra-low latency) achieved by bringing AI services closer to end users.
Artificial Intelligence, Real Innovation
With AI baked directly into 6G, it will be possible to intelligently connect distributed intelligent agents, driving large-scale deployment of AI in all industries. Spectrally efficient, high-capacity, and low-latency transmission for distributed learning — including data and model parameter exchange among large numbers of intelligent agents — is expected for real-time AI. And native trustworthiness, with the support of native security and local data privacy, will be a key enabler for achieving this.
For AIaaS, 6G functions as a native intelligent architecture that deeply integrates communication, information, and data technologies, as well as industry intelligence, into wireless networks, serving all types of AI applications with large-scale distributed training, real-time edge inference, and native data desensitization.
In today’s networks, AI services are located in a central cloud at the application layer. This will change with 6G networks, where the network architecture and AI will go hand in hand. Put differently, native AI support will be one of the fundamental factors that drive innovation in the network architecture.
So while many of us are already familiar with AI enabling things like self-driving cars and even smart toothbrushes, look out for when 6G comes to market, as the combination of 6G and AI will enable untold possibilities.
Subscribe to this blog to keep pace with this series on 6G – as well as the latest tech – and download the white paper: 6G: The Next Horizon – From Connected Things to Connected Intelligence.
Disclaimer: Any views and/or opinions expressed in this post by individual authors or contributors are their personal views and/or opinions and do not necessarily reflect the views and/or opinions of Huawei Technologies.