IMMERSE6G - AI/ML high performance computing for wireless 6G in the Metaverse

The benefits of artificial intelligence (AI) and machine learning (ML) are impacting all sectors of society and have become a key area where a great deal of experience and knowledge is needed to master all the details. Recently, the concepts of digital twin and metaverse have further enhanced the use of computation to manage complex day-to-day processes, measure the impact of current decisions in the near future or facilitate the design of new communication protocols. However, the more complex the problem to be addressed, the more computational capacity is required. This capacity is very expensive, so the availability of these types of infrastructure is generally scarce. IMMERSE6G will provide an infrastructure capable of: i) AI-intensive computing with powerful state-of-the-art GPUs, ii) non AI-intensive computing, with (lower power) GPUs, CPUs and FPGAs, to be able to perform general and edge computing, iii) to allow working in the metaverse and digital twin. The equipment considered in IMMERSE6G will allow the development of the research lines on the objectives: o1) disruptive enabling techniques for learning and computing, o2) implementation of the digital twin and use of the metaverse for 6G, o3) development of machine learning algorithms in distributed environments, and o4) development of tools for estimating energy consumption of 6G applications, carbon footprint computation and deployment of mitigation strategies, as well as o5) micro-orchestration of RAN resources for network optimization and energy efficiency . The architecture to house the equipment has been sized to be scalable in the future and easy to increase the computing capacity. It is expected to increase the number of research lines thanks to this infrastructure, at a time when it is essential to have the best technology available to work in AI and 6G in order to deploy research at the right speed and be competitive worldwide.