Control and learning methods for cognitive spaceborne tracking radars

Making radio frequency remote sensing technologies fully autonomous enables the acquisition of more insightful data during emergency situations. This thesis investigates the application of control techniques and Artificial Intelligence (AI) to cognitive radars. This class of radar systems utilizes information about the surrounding environment, extracted by the receiver, as feedback to the transmitter, thereby rendering the radar a closed-loop system. Within this framework, the aforementioned algorithms can leverage that feedback to adapt the transmission configuration of the radar payload. In this work we mainly focus on cognitive tracking radars. Firstly, we propose an optimal-control algorithm for a spaceborne cognitive tracking radar. The algorithm computes the optimal sequence of transmitted waveforms during tracking, thereby minimizing the estimation error of the trajectory of the target of interest. Simulation results demonstrate that this approach outperforms other methods reported in the literature. Subsequently, the focus shifts to a Tip & Cue configuration for maritime surveillance. This framework involves the cooperation of two satellites to enable higher-resolution data acquisition over an area of interest. Initially, we focused on developing a Gaussian Process Regression method to reconstruct vessel trajectories from sparsely sampled Automatic Identification System (AIS) data. AIS is used by cooperative ships to transmit information such as position, speed, and heading. The algorithm was subsequently implemented and tested on space-qualified Field-Programmable Gate Arrays to evaluate its on-board feasibility. Next, we focused on the development of a target-tracking algorithm based on Synthetic Aperture Radar (SAR) detections. The model employs a combined target-dynamics representation, e.g., partly model-based and partly data-driven (trained via a neural network), to predict the target state, and subsequently refines the estimate using AI-based SAR detections. Subsequently, we extended this problem to a Synthetic Aperture Radar with the objective of detecting and tracking vessels in open sea. In this setting we proposed a Tip & Cue framework, which involves the cooperation between two satellites to enable more accurate imaging of an area of interest. Sensor fusion between Synthetic Aperture Radar and the Automatic Identification System served as an enabling technology for this use case, allowing the association of radar-detected vessels with AIS messages received by the onboard satellite receivers.