Self-Optimizing 6G RAN: A Unified Approach with Image and Video Analytics

The advent of 6G wireless technology heralds a new era of network optimization driven by AI/ML algorithms across its entirety. As outlined in 3GPP TS 28.104 (Release 19), Management Data Analytics (MDA) presents diverse use cases and requirements. However, the variety of data formats and AI/ML application methods introduces complexity and limits generic implementation. Compounded by the exponential deployment of cells and a range of new devices like XR devices, UAVs, etcs., the resulting data deluge poses a formidable big-data challenge for 5G advanced, 6G, and beyond. Anticipating the unique demands of 6G, characterized by metaverse-enabled networking and AR/VR applications, there arises a need for AI/ML-driven visual data processing services. This paper presents a novel approach: representing wireless network data in the form of images and videos. Such a representation not only optimizes storage and processing requirements but also facilitates the reuse of AI/ML models across various use cases and data formats. Experimental validation is provided through two key avenues: Firstly, text-based logs are transformed into images, enabling the identification of recurrent issues through visual pattern recognition. Secondly, Comma Seperated Variables (CSV) data is converted into a series of images, effectively simulating video frames, thus extending the realm of video analytics to encompass all data types pertinent to self-optimizing networks in 6G RAN. The advantages of this image and video-based data representation paradigm are meticulously elucidated, offering a compelling proposition for the future landscape of wireless network optimization in 6G and beyond.