Performance Prediction for Dependable 6G Networks through Causal Artificial Intelligence

Purpose and goal

6G networks are supposed to provide novel services towards the cyber-physical continuum. This comprises communication and compute workloads, with a high level of dependability towards real-time sensitive tasks. This project has the goal to support such service provisioning in 6G networks by enabling accurate predictions of future state evolution. Our novel angle to this approach is to leverage causality, which is a statistical theory of dependent states, in AI algorithms.

Expected effects and result

Causal AI provides is a powerful tool whenever basic data is expensive or difficult to obtain. For 6G networks, one important problem is the prediction of rare events that lead to significant performance degradation, which is important for applications of the cyber-physical continuum. We strive to develop algorithms and methods that help to predict such rare events with as little data as possible. For that, we will provide algorithms for causal discovery as well as causal AI approaches.

Planned approach and implementation

The project has three main parts. In the first one, we will extend existing prototypes of wireless networks towards data collection. Having a substantial amount of data is important for the later project stages. In stage 2, we will work on different methods for causal discovery, given the data we obtained in the first stage. Finally, we will develop causal AI methods for performance prediction of 6G networks.