Core Concepts
Proposing a novel method using generative deep learning to create low probability of detection radar waveforms that blend into the RF background while maintaining sensing capabilities.
Abstract
Radar systems are crucial for military operations, but vulnerable to detection.
Low Probability of Detection (LPD) radar waveforms aim to reduce detectability while maintaining performance metrics.
Traditional LPD waveform design methods include low peak power, wideband, frequency agile, and coded waveforms.
Deep learning techniques like GANs are used for LPD radar waveform design.
Unsupervised GANs generate adaptive radar waveforms matching the RF environment distribution.
Wasserstein GANs optimize waveform similarity to RF background for LPD.
Ambiguity loss optimizes generated waveforms for desirable ambiguity characteristics.
Evaluation shows up to 90% reduction in detectability with improved ambiguity function characteristics.
Stats
Our method can generate LPD waveforms that reduce detectability by up to 90%.
Quotes
"Our framework also provides a mechanism to trade-off detectability and sensing performance."