By Sarah LaSelva, Director of 6G Marketing at Keysight
The artificial intelligence (AI) revolution has arrived. With the public release of apps like ChatGPT, people are able to get hands-on experience using the power and potential of deep neural networks and machine learning (ML). ChatGPT is a language model that is trained using massive amounts of text data from the internet and books to generate texts similar to those written by real people. This type of application perfectly demonstrates the advantages of artificial intelligence. It can continuously optimize the output in complex scenarios through a large amount of training data.
Wireless networks are inherently complex, generating large amounts of data, and their complexity continues to increase with each new generation of new technologies introduced. These features make AI an ideal tool for optimizing wireless networks.
Application of AI in 5G networks
As 5G technology matures, AI and ML have been introduced into research by 3GPP (Third Generation Partnership Project), an international standards organization that sets standards for cellular technology. Artificial intelligence is currently being considered to improve the air interface, including network energy saving, load balancing, and mobility optimization. Due to the large number of potential use cases for air interfaces, only a small subset was selected for investigation in the upcoming release of 3GPP R18, covering channel state information (CSI) feedback, beam management, and positioning. It is important to note that 3GPP does not develop AI/machine learning models. Instead, it seeks to create a common framework and evaluation methodology for deploying AI/ML models into different functions of the air interface [1].
In addition to 3GPP and air interfaces, the O-RAN consortium is exploring how to leverage AI/machine learning to improve network orchestration and management. For example, the O-RAN Alliance's architecture has a unique feature, called the RAN Intelligent Controller (RIC), which is primarily used to assist AI and machine learning in optimizing different use cases. RIC can manage both near real-time applications (xApps) and non-real-time applications (rApps). xApps for improving spectrum efficiency and energy efficiency, and rApps that leverage artificial intelligence for network orchestration and management already exist. As the O-RAN ecosystem grows and matures, more xApps/rApps will emerge as well as applications that leverage RIC-based AI and machine learning optimization.
Figure 1: ORAN network
6G network native AI technology
Although 6G is in its infancy, what is certain is that artificial intelligence/machine learning will become a fundamental part of all aspects of future wireless communication systems. At the network level, although there is no formal definition, the term "AI native" is already widely used in the industry. One way to look at these AI-native networks is to extrapolate the above diagram based on current virtualization technologies and deaggregation trends in RAN (Radio Access Network) (Figure 1). Each chunk in the network may contain AI/ML models, which can vary between different vendors and applications (Figure 2).
Figure 2: ORAN 6G network
AI-native networks can also be used to refer to networks built to run native AI/machine learning models. Refer to the design flow below (Figure 3). In a traditional 5G network, the air interface is made up of different parts, each designed by a human. In the 5G-Advanced network, each segment will utilize machine learning techniques to optimize specific functions. In a 6G network, it is possible for artificial intelligence to design the entire air interface using deep neural networks.
Figure 3: Development from integration with AI to AI-native networks[2]
AI/ML optimization
Drawing on the idea that AI/ML can be used to improve network orchestration and management, 6G hopes to leverage AI and machine learning to solve optimization challenges. For example, AI can turn components on and off based on real-time operation to reduce power consumption across the network. Today, xApps and rApps achieve this at the base station level by turning energy-hungry components such as non-operating power amplifiers on and off. However, AI's ability to quickly solve challenging computational problems and analyze massive amounts of data makes it possible to optimize network performance on a larger scale or across the city, or nationwide. The entire base station can be shut down during the period of low frequency of use, or the cell can be reconfigured to use as few resources as possible to meet the real-time needs of users in a green, low-carbon, energy-saving and environmentally friendly manner. It is not yet possible to reconfigure the base station and the entire city's network in this way, and reconfiguring and testing any changes to the network configuration often takes days or weeks. Nevertheless, the development prospects of different AI technologies are very promising, and they remain a top consideration for infrastructure providers.
summary
The application of artificial intelligence in wireless networks will not wait until the advent of 6G networks. Active research is underway throughout the ecosystem to develop new models and integrate these models into existing and future wireless communication systems. However, these models are still new and need to be evaluated for rigor and reliability. Properly training AI models on different datasets, quantifying their improvements over traditional technologies, and defining new testing methods for AI-driven modules are all critical steps that must be taken as new technologies are adopted. As AI models and test methods and technologies mature, there is no doubt that AI will revolutionize the wireless communication industry in the next 5-10 years.