ExpectedOutcome:
Proposal results are expected to contribute to one or more of the following outcomes:
- Development of European processors/accelerators for secure embedded intelligence supporting inference and machine learning in autonomous intelligent objects and AIoT (Artificial Intelligence of Things) at the edge.
- Development of platforms and automated tools for HW/SW co-design of edge-AI based systems including non traditional computing paradigms (e.g. neuromorphic).
- Development of AI software algorithms and/or frameworks and tools for dealing with sparse computing resources for edge-AI based systems.
- Increased energy efficiency and improved processing capabilities for edge-AI.
- Support of virtualization concepts at firmware, operating system and middleware layer for vertical applications (e.g. automotive, industry, … )
- Design for security and reliability to support applications in critical sectors like mobility, healthcare, industry, …
Scope:
Proposals need to address:
- The design of innovative and energy-efficient processing solutions for AI on theedge and deep-edge, with a focus on new processor architectures and middleware.
- Tools allowing semi-automatic and automatic design space exploration, including variants of algorithms, computing paradigms, hardware performances, energy efficiency, etc.
- Scalable architecture, use of interposer and chiplets to build chips for various applications (for edge and for embedded HPC applications) with the same family of hardware building blocks with efficient interconnection network, e.g. using photonics.
- Development of hybrid architectures, integration and cooperation of neuromorphic or other non-conventional computing solutions within classical systems. Supporting design tool chains and OSs addressing multiple computing paradigms
- Middleware and engineering tools, to reach a trade-off between training cost, power consumption and execution time while supporting virtualization concepts.
- Advanced memory management
- Automated transfer-learning, meta-learning, and real-time learning at the edge
- Secure (i.e. trustworthy and explainable) edge-AI by design.
- Encouraging SMEs to participate in those developments, in particular paying attention to the needs of SMEs, involve SMEs in project execution, and develop solutions that can be taken up and/or exploited by SMEs
Specific Topic Conditions:
The activities have their centre of gravity at the TRL 4-5 – see Annex 1 of the KDT JU Work Pogramme 2021
