Intelligence on the Edge

Strategic Research Area 3

Distributed Intelligence

Guided by Associate Professor Miguel Bordallo López, this strategic research area recognises data as the cornerstone of the future digital society. The anticipated massive data streams resulting from the ultra-densification of communications and sensing pose a significant challenge to the existing data and computing infrastructures.

In more detail, the onslaught of generated data calls for distributing computations over the device-edge-cloud computing continuum. Along the continuum, novel distributed paradigms for AI/ML model training and inference are required, leveraging the capabilities of advanced 6G networks to enhance privacy and data locality. Distributed computing and AI/ML will play pivotal roles in enabling distributed sensing, modelling, and the development of unique user interfaces.

These advancements will need integration into complex system architectures that operate independently and in dynamic contexts, thereby laying the solid foundation needed to meet the future society’s demands.

Distributed AI

Future data-intensive applications need distributed AI and analytics solutions on the edge and fog computing platforms. With the advent of 6G, these solutions will help complement human decision-making, construct autonomous systems from small devices to whole factories, and optimize network performance and marshal billions of devices.

Achieving varied service and latency needs while increasing resource efficiency is the key challenge of future AI. Updated application-level protocol and security solutions are required.

We develop ways to maximize AI method distribution across heterogeneous nodes. Increasing heterogeneous streaming data will demand sophisticated computing paradigms and distributed and self-organizing methods. Furthermore, achieving system-level goals and efficiently employing resources needs system models for decision making.

Multimodal sensing and modelling

6G relies on multimodal sensor data to detect and model the surroundings. New sensors and actuators accompanied with high-speed connectivity and low-cost computational processing have made real-time and distributed intelligent applications feasible.

The challenge is to make sense of all the data. Uncertainty quantification and propagation are automated ways to improve operational data quality and privacy. It is possible to increase the trustworthiness of smart decision support systems by improving data visibility from several sources and understanding the functions and logic behind the judgments.

6G is intended to natively allow radio-based sensing and support ultra-dense sensor and actuator networks, enabling hyper-local and real-time sensing, communication, and interaction. Both the physical and programmed worlds require multidimensional orchestration.

Distributed INTELLIGENCE

Publication Highlights

Ding, A. Y., Peltonen, E., Meuser, T., Aral, A., Becker, C., Dustdar, S., … Wolf, L.
Roadmap for edge AI: a Dagstuhl perspective

Elgabli, A., Park, J., Bedi, A. S., Bennis, M., Aggarwal, V.
GADMM: Fast and Communication Efficient Framework for Distributed Machine Learning

Lähderanta, T., Leppänen, T., Ruha, L., Lovén, L., Harjula, E., Ylianttila, M., … Sillanpää, M. J.
Edge computing server placement with capacitated location allocation

Bhayani, S., Sattler, T., Barath, D., Beliansky, P., Heikkilä, J., Kukelova, Z. 
Calibrated and Partially Calibrated Semi-Generalized Homographies

Niu, X., Yu, Z., Han, H., Li, X., Shan, S., Zhao, G.
Video-based remote physiological measurement via cross-verified feature disentangling

SEE THE ACTION

Sample projects

ILLUSIVE: Foundations of Perception Engineering

FRACTAL: Empowering Edge Computing

MAALI: Multisensory automation for assisted living

Frages: A framework for General Spatial AI

R2D2: Robust and Reliable Delay-sensitive Communication and Control Co-design