Trustworthy Edge Artificial Intelligence
| Reference number | |
| Coordinator | Kungliga Tekniska Högskolan - KTH Skolan f elektroteknik och datavetenskap |
| Funding from Vinnova | SEK 200 000 |
| Project duration | July 2026 - August 2026 |
| Status | Ongoing |
| Venture | 6G - Competence supply |
| Call | Individual mobility within 6G for collaborations with the USA and Singapore |
Purpose and goal
This project aims to develop Edge AI algorithms that are trustworthy, energy-efficient, fair, and resilient for sustainable and inclusive intelligent systems. Future 6G applications, including autonomous mobility, immersive communications, digital twins, and smart health, will rely on distributed intelligence across resource-constrained edge devices. However, current Edge AI methods overlook energy use, fairness, robustness to unreliable devices, and trust under uncertainty.
Expected effects and result
The project will design communication-efficient and decentralized learning algorithms that reduce energy and bandwidth costs while preserving performance. It will introduce fairness-aware mechanisms to avoid disadvantaging users with diverse data distributions or connectivity conditions. It will also develop resilient methods that tolerate quantization, intermittent communication, non-iid data, and faulty or adversarial nodes.
Planned approach and implementation
The project is implemented in three stages. First, eligibility requirements from 6G applications and constraints of edge devices are mapped, focusing on energy, bandwidth, fairness and robustness. Then, decentralized learning algorithms and mechanisms for trust, fairness and resilience are developed. The methods are evaluated in simulated and realistic scenarios with non-iid data, disturbances and unreliable nodes. The results are disseminated through reports and dialogue with relevant actors.