Characterization and recognition of GNSS interference through federated learning (DARCII)

Federated Learning for (mobile) GNSS-interference Analysis:
The overall goal of this project is to orchestrate current AI technologies for personalized jammer characterization across (geographically) distributed and mobile detection nodes. The aim is to achieve resilience to different environments, antenna patterns and unknown jammers. Federated learning of GNSS interference at the sensor node to reduce local influences and in the cloud to adapt a device-specific interference pattern as a fingerprint are being researched and developed for this purpose. By learning with a few reference points (Few-shot Learning), the AI models are adapted to novel interference patterns. Additional information about the fingerprint, such as an image, is also stored manually. The aim is to display the relevant result information in a web frontend. To orchestrate the AI technologies, a decentralized DARCII data space is being set up and its opening for external data is being developed. By integrating these external providers, a high degree of generalizability of the AI models is achieved. This enables scientific and technical progress compared to third-party providers.

More information

 

GNSS receiver technologies

Satellite navigation receivers and antennas for a wide range of applications

 

Data Analytics and Machine Learning

Robust tracking algorithms and data analysis methods using machine learning and statistical methods