Standardization and technology trends of artificial intelligence for mobile systems

Today, the applications of AI/ML techniques are so pervasive and numerous that they cannot be easily listed. However, some of the most successful and well-known applications include search engines, targeted advertisements, recommendation systems, autonomous vehicles, language translation, image recognition, and large language models. Although introduced more recently, AI/ML applications for mobile systems have already shown potential in improving performance and user experience, automating network management and optimization, reducing overhead and cost, among others. This paper looks into the standardization activities in 3GPP and O-RAN for AI/ML applications, and future technology trends of AI/ML. Though still in its early stages, 3GPP has started to incorporate intelligence into the core network, RAN, and the air-interface. Meanwhile, O-RAN has introduced SMO and RIC for RAN automation and intelligence. Even for the air-interface, where near real-time operation is required, 3GPP is exploring AI/ML use cases in areas such as beam management, CSI feedback enhancements, and positioning, to improve performance and reduce overhead. Federated learning, E2E learning, and explainable AI/ML are emerging as important technology trends in the field of AI/ML for the future. When these future AI/ML technologies are integrated with mobile systems, it is expected that a series of cross-layer protocol stacks for communication could be replaced by an appropriate AI/ML model that can provide more automated, reliable, and accurate inference and actions compared to traditional methods.

Automated summary

The applications of AI/ML techniques are pervasive and numerous, including search engines, recommendation systems, autonomous vehicles, etc. In mobile systems, AI/ML has the potential to improve performance, automate network management, and reduce costs. Standardization activities for AI/ML applications are being conducted by 3GPP and O-RAN. Future technology trends include federated learning, E2E learning, and explainable AI/ML.