The expected outcomes are addressing several technologies and architectures that have the potential to simplify the operations of a network, decrease operational expenditures, or allow for new service or service models to be deployed. They target mainly:

• The definition and design of innovative more efficient and simplified advanced 6G architectures (Control and/or User plane) enabling a seamless grow of service deployments without constraints, targeting long-term evolution of 6G architectures.
• The definition and performance characterisation of novel techniques for constrained AI operations, notably from an energy perspective, from real-time learning perspective and for optimal deployment of computation resources.
• Identification and performance characterisation of technologies and architectures for provision/orchestration of a multiplicity of telco or verticals backend services through serverless computing across multiple domains and for vertical use cases.

Objective:

Please refer to the "Specific Challenges and Objectives" section for Stream B-01 in the Work Programme, available under ‘Topic Conditions and Documents - Additional Documents’.

Scope:

Proposals should address one of the following five areas (and clearly identify which area is being addressed):

1. New architectural solutions: New architectural solutions targeting the simplification of the architecture (user and control plane) enabling operators to introduce and operate services more efficiently, effortlessly integrate and connect new network domains (considering that any device can operate as a network node), as well as enable users to seamlessly roam across operators, and network technologies and domains. It covers current limitation of the Service-Based Architecture (SBA), legacy constraints and evolution from early 6G architectures (e.g. Hexa-X II model).
2. Deep learning models: The development and evaluation of deep learning models that include predefined constraints either from a network operation viewpoint or from a user service provision viewpoint. The constraints may be in the form of physical law, logical rule, or any other domain specific knowledge, with an end-to-end connectivity perspective. Of particular interest are problems related to AI optimization under energy and/or under security constraints. The scope covers constrained deep learning by incorporating constraints into the learning process for network processes optimisation, as well as interconnection of untrusted data sources.
3. Real time serverless computing: The scope covers provisioning and orchestration algorithms and technologies for database and storage services towards the widespread implementation of “Function-as-a-Service” (FaaS). It allows execution of code on various parts of the network (e.g. edge) and support the versatile/optimised function placement and dynamic replacement expected in 6G. Also in scope is research on methods for instant start (to overcome the cold start characteristics of the scheme), high-dimensional task orchestration across multiple stakeholders, and imposing embedded security in such infrastructures.
4. Autonomous Cognitive Agents: The definition of architectures with a focus on simplicity, scalability, and security with a focus on the holistic combination of service composition and knowledge handling in a network-compute continuum where the network and applications merge in an organic way and operations are carried out in a decentralized manner. These Multi-agent systems, possibly but not necessarily based on LLMs, should be able to excel the performance of individual agents. The agents can potentially invoke each other spontaneously and can be operated in a decentralized manner, thus departing from the relatively static architectures we know today, but will need to remain under effective control of adequate service management models.
5. Goal-oriented Communication: The definition and design of revolutionized effective goal-oriented communication protocols, languages, and media among devices and machines equipped with AI, especially those with generative AI technologies, allowing them to extract and communicate only goal-relevant information, possibly directly in various waveform formats, to reduce communication, computing, storage, and energy consumptions. Full system concepts are expected, including the ancillary control and management aspects imposed by these technologies.

Note: To ensure a balanced portfolio within this cluster of activities focusing on Advanced Architectures Systems and Technologies, grants will be awarded to proposals not only in order of ranking but at least also to one project that is the highest ranked within each of the above 5 areas/priorities, provided that the proposals attain all thresholds (and subject to available budget). Proposals may want to address several of the proposed priorities, but they should indicate clearly what is the centre of gravity of their proposal (i.e. the main covered priority). See General call conditions, section 3.F of Appendix 1.