Quantum Reinforcement Learning for Cognitive SAR

Abstract

Cognitive radar involves a closed loop between the radar receiver, radar transmitter, and the environment, similar to the perception-action cycle in human cognition. It adapts acquisition parameters based on the acquired information, which offers more efficient resource management of future SAR missions in varying scenes. We propose a spaceborne cognitive SAR concept for ship detection, which uses a two-stage process to improve ship detection capabilities. This radar concept can be implemented on a single platform using a hybrid mode or distributed on separate SAR satellites operated in a convoy configuration. We implement variational quantum circuits for the adaptation to the scene in a reinforcement learning approach to explore possible advantages of quantum computing combined with parameter optimisation on classical computers.