ARIST-6G: Adaptive Receiver via Interpretable Sequence Transformation for 6G Networks
| Reference number | |
| Coordinator | Kungliga Tekniska Högskolan - Avdelningen för teknisk informationsvetenskap |
| Funding from Vinnova | SEK 2 550 000 |
| Project duration | April 2026 - August 2027 |
| Status | Ongoing |
| Venture | 6G - Competence supply |
Purpose and goal
Objectives are developing theoretical foundations linking ICL to classical estimators, rapid and efficient decoding, and intelligent pilot selection. ARIST-6G proposes an interpretable Transformer based In-Context Learning receiver that learns directly from pilot symbols without explicit CSI. ARIST-6G advances the state of the art along three dimensions, i.e., theoretical interpretability, computational efficiency and robust adaptability.
Expected effects and result
ARIST-6G advances the state of the art along three dimensions. First, theoretical interpretability is established by rigorously proving correspondence between linear ICL and Bayesian estimators, including error bounds and approximation analysis. Second, computational efficiency is achieved with scalable quasi-linear per-symbol processing. Third, robust adaptability. Results include publications, trained researchers and knowledge of 6G receivers.
Planned approach and implementation
The project plans to apply advanced AI and signal processing tools to develop ICL receivers in 6G environments. Extensive simulations will be used to verify theoretical results.