Fraunhofer Institute for High Frequency Physics and Radar Techniques FHRCognitive radar systems respond to their environment and adjust their parameters in real time. Fraunhofer FHR is contributing to the development of a demonstrator with space-capable hardware that is intended to illustrate these capabilities. The institute provides a pre-trained neural network as well as decision criteria to detect and avoid disturbances in a radar signal before the SAR (Synthetic Aperture Radar) image is captured. SAR satellites deliver high-resolution images of the Earth's surface day and night. So far, raw data are mostly transmitted to Earth and processed and analyzed on the ground with classical algorithms.
New systems enable onboard data processing and adaptation
Modern onboard computers now open up a new class of “intelligent” instruments: cognitive radar systems that autonomously adapt their operation to the current situation.
A cognitive radar observes the scene, evaluates the received signals, and adjusts its mode of operation in real time, for example by:
- Changing frequency, pulse width, or transmit power
- Adjusting the orientation and shape of radar beams (beamforming)
- Switching image modes (e.g., broad overview vs. small, high-resolution excerpts)
- Interacting in formations with other satellites or the platform itself.
Crucially, this is done directly on board within a closed feedback loop, without the need for a ground operator.
Artificial intelligence expands the radar system’s capabilities
Machine Learning (ML) and Deep Learning have been successfully used for several years on the ground to analyze earth observation data, for example for object detection, environmental and disaster monitoring. The potential of AI within the radar system itself remains largely untapped. This is where the researchers focus: investigating how ML can be used directly in the radar signal chain to enable truly cognitive SAR. The AI analyzes raw or preprocessed SAR data on board and decides how the system should optimally use its resources—transmit power, bandwidth, observation mode, data rate. For operators this means higher data quality with more efficient use of scarce mission resources.
Cognitive SAR solution for interference suppression
As a use case, the suppression of disturbances (Radio Frequency Interference Mitigation, RFI) in operational use was chosen. These disturbances in frequency bands and radar pulses from SAR satellites caused by external sources are increasingly occurring and can render SAR images partially or completely unusable.
The project aims to develop a cognitive SAR solution that autonomously detects, assesses, and effectively suppresses RFI on board. The approach consists of:
1. Real-time detection: ML models analyze the received raw data and identify typical interference patterns, such as narrow spectral lines, impulsive signals, or recurring patterns in certain frequency and time ranges.
2. Autonomous adaptation (Closed Loop): Based on this assessment, the radar adjusts its operating parameters, for example by shifting the used bandwidth.
Thus the full capability of the SAR instrument is leveraged where it is actually used, and users benefit from robust, high-quality data sets and increased mission safety—even in the face of growing competition in the frequency spectrum.
In the ESA-funded project “Closed Loop Artificial Intelligence Cognitive Synthetic Aperture Radar,” Fraunhofer FHR works with the German Aerospace Center (DLR) as a subcontractor for Airbus.