About Us
Our proposed Nordic university collaboration is organised around three main objectives:
Joint research on key aspects of Edge Intelligence
Use case studies and outreach to stakeholders
Training, networking and dissemination.
Joint Research Activities:
Several central challenges must be addressed jointly to accelerate the necessary research towards massively scalable edge intelligence systems. We see in particular three domains to be of key importance:
(a) Edge Architectures: Our main objective regarding architectures is to enable edge intelligence for continuous and heterogeneous data streams from, e.g., the Internet of Things (IoT) or extended reality (XR) devices. In the NUEI project, we will develop edge intelligence architectures that can handle vast amounts of heterogeneous data and provide services with high performance and quality. Central aspects impacting future architectures beyond scalability are energy efficiency, locality, latency, and accuracy.
(b) Distributed / Hierarchical Inference: Scale in model size implies increasing constrained placement of models to be traded with other aspects, such as bandwidth, latency, and potentially locality. It is foreseeable that large-scale models will increasingly have to be distributed over multiple locations to achieve placement objectives. However, the distribution of large-scale models and the involved costs and overheads, imply trade-offs that are not well understood today. This also applies to hierarchically staged approaches to inference with increasing model complexity and accuracy. How to devise stages and operate (i.e., offload) inference tasks between the stages is far from understood today.
(c) Semantic communications: is regarded as a disruptive technology for next-generation communication systems as it promises to open the path towards time-critical and resource-demanding applications, such as high-resolution 3D video streaming or holographic communication.
Use Cases and Outreach:
Beyond the above-mentioned research directions in Edge Intelligence, the consortium sees an important further objective in the joint development of applications and use cases. Different competencies in the consortium, as well as different technical/testbed capabilities, provide several opportunities.