Project results are expected to contribute to the following expected outcomes:

• Novel AI-enabled Cloud and Edge management solutions tailored for the processing needs of AI workloads across the cognitive cloud-edge-IoT continuum.
• Strategic industrial cooperation across the Cloud-Edge-IoT cognitive computing continuum to support future hyper-distributed AI applications.
• Seamless and trustworthy integration and interoperability across diverse computing and data environments spanning from core cloud (including HPC) to edge to IoT and across different technology stacks.
• Enhanced openness and open strategic autonomy in the evolving data and AI-economies across the computing continuum validated through key business/societal sectors.
• Guaranteeing a minimum level of interoperability and portability thereby facilitating European access to foreign markets.

Scope:

The Cloud to Edge Continuum needs to provide seamless and trustworthy integration of diverse computing and data environments spanning from core cloud to edge to IoT and support the enormous data, processing needs, and new resource types brought by next generation AI technologies.

Different types of AI processes pose different requirements that compute infrastructures need to meet to execute them. The state-of-the-art in generative AI and large language models is heavily reliant on high-performance processing and very large AI models. Cutting-edge hardware accelerators that power these processing systems are scarce on the market and only available in highly specialised, high-performance infrastructures in certain cloud and HPC environments at considerable costs. At the same time, the requirement to gather, process, and transmit massive amounts of data to the central data processing environment remains a barrier for many AI applications. All these factors urge the emergence of efficient tools and mechanisms to empower the distribution of AI training and inference processes throughout the computing continuum.

Empowering the next generation AI technologies with on-demand, agile and situation-aware infrastructure that brings data- and computing power to where and when it is needed will let end-users exploit Artificial Intelligence across the computing continuum without compromising on security and trust and optimising their energy use. These challenges span various aspects of the continuum, including on-device data processing, data orchestration and sharing, AI integration, decentralised intelligent management, decentralized and global optimization, energy and resource heterogeneity support, data management, security/privacy, and synergies with 5G/6G. Addressing these challenges is crucial for realizing the vision of a cognitive cloud-to-edge continuum as a key enabler for any emerging trends such as AI/generative AI.

The Cognitive Computing Continuum could eventually be extended to include other computational resources, such as HPC, and provide abstraction layers to maximize the benefits of available hardware.

Addressing all the above complexities calls for innovative research to overcome these challenges. The aim is to develop generic and AI-enabled cloud-edge technologies encompassing the whole computing continuum to empower the development of AI/generative AI technologies and applications. The proposals should demonstrate the generic applicability of the proposed technological solutions across various application domains such as but not limited to, manufacturing, healthcare, robotics, transportation and smart cities.

The following (one or more) research areas should be addressed:

• Development of novel mechanisms for the efficient development, deployment, and operation of AI workflows across heterogeneous and distributed infrastructures along the Edge to Cloud to HPC continuum that optimise training times, model accuracy and data management while factoring in performance metrics such as memory usage, energy efficiency, application processing and data transfer latency, and network overheads. These should factor in virtualisation and orchestration techniques that seamlessly integrate heterogeneous processor architectures and cater for the explainability of the applied cognitive optimisations.
• Decentralised and federated computing continuum tools and mechanisms to enable distributed AI architectures. These include scheduling, orchestration, and placement mechanisms that leverage the wide range of Edge computing environments available in the compute continuum, including on-device edge. Tools and mechanisms should take into consideration - where appropriate - data security and privacy aspects. The focus is on enhancing AI process execution through techniques such as model, data, hybrid parallelism and data compression, gossip, swarm, and federated training, or conditional computing.
• Cloud and edge processing tools and techniques to reduce AI processing power usage and emissions across the cognitive computing continuum, relying on hardware efficiency (for example, thanks to special-purpose accelerators and heterogeneous hardware processor architectures) and energy optimisation techniques, such as hardware and software approximation.

This topic implements the co-programmed European Partnership on AI, Data, and Robotics.

Projects are expected to develop synergies and relate to activities and outcomes of the Digital Europe Programme (DEP) and any existing or emerging Important Projects of Common European Interest (IPCEI) initiative, such as IPCEI-CIS.

All proposals are expected to share communicable results with the European R&D community, through the AI-on-demand platform, and if necessary other relevant digital resource platforms in order to enhance the European AI, Data and Robotics ecosystem through the sharing of results and best practice.