Fraunhofer Institute for Integrated Circuits IIS
Use of AI for Optimized Angle Estimation and Interference Source Localization
The PaiL project aims to research and develop AI-based methods for precise angle resolution and robust localization of multiple interference sources under realistic multipath conditions. The goal is to significantly improve the accuracy and speed of classical direction-finding methods and to expand their capabilities in complex environments. The use of neural network models aims to achieve greater robustness against signal noise, reflections, and model-dependent estimation errors.
In a step-by-step approach, classical super-resolution methods such as MUSIC, ESPRIT, and SAGE are analyzed and supplemented with AI-based approaches. Building on the characterization of relevant environmental and signal parameters, suitable neural models are implemented to optimize angle and source estimation. In addition, time-sensitive methods—including Kalman filters, LSTM networks, and Transformer-based models—are employed to adaptively map dynamic signal trajectories.