Recurrent Waveform Optimization for Desired Range-Doppler Profile With Low Probability of Interception: A Particle Filter Approach

Modern cognitive radars, armed with knowledge-aided waveforms, exhibit considerable effectiveness in detecting low-speed, small radar cross-section (RCS) targets, while demonstrating resilience in electronic countermeasure scenarios. In this article, we introduce a novel technique named particle filter-based recurrent waveform optimization (ALTERATION) to design unimodular waveforms with tailored range-Doppler profiles and low probability of intercept (LPI), thereby enhancing radar performance in complex environments. ALTERATION leverages the power of particle filters (PFs) in conjunction with iterative optimization methods. Our approach begins with linear frequency modulation (LFM) waveforms regularized by chaotic-phase terms serving as the initial particles. During each PF iteration, these particles are evolved via an alternating optimization method dovetailed with fast Fourier transform (FFT) operations. A specially proposed perturbation scheme corresponding to the deployed optimization method is then applied to weigh particles according to their auto-correlation levels, facilitating resampling to produce the new particle set for the next PF iteration. Moreover, ALTERATION is designed to readily integrate low-resolution phase quantization, thereby further mitigating hardware costs while maintaining near-optimal waveforms. Numerical comparisons illustrate that our ALTERATION approach outperforms existing state-of-the-art alternatives, indicating its potential for application in cognitive radar systems.