Performance and Inference Time Tradeoff for RNN Model based Wideband Inter-radar Interference Mitigation

With recent advancements in autonomous driving, there has been growing interest in using CS (Chirp Sequence) radar for precise sensing. Unlike other in-vehicle sensing techniques, CS radar offers cost-effectiveness, compactness, and greater resilience to environmental influences. However, the widespread adoption of CS radar may lead to radar-to-radar interference, causing issues like target non-detection and false detection. To tackle this challenge, a method utilizing RNN (Recurrent Neural Networks) for interference mitigation has been proposed. This study specifically investigates VanillaRNN and biLSTM models based wideband interference mitigation methods and explores their tradeoff between mitigation performance and inference time.