Radar Waveform Design Method Based on Cascade Optimization Processing under Missing Clutter Prior Data

Cognitive radar waveform design often relies on accurate clutter prior information. When prior information data is missing, the constructed clutter model will be severely mismatched, affecting the radar’s ability to suppress clutter. Aiming at the radar waveform optimization problem under missing clutter prior data, this paper establishes point and block-like missing scenarios under the completely random missing mechanism, designs a waveform optimization model with constant modulus and similarity constraints, and proposes a radar waveform training algorithm based on priority filling−reinforcement learning cascade optimization: that is, a cascade method in which the reinforcement learning agent interacts with the clutter environment repaired by a filling algorithm, with the optimization goal of maximizing the signal-to-noise ratio, and the optimal configuration strategy with waveform parameters is obtained through iterative training. Finally, simulations verify the superiority of the proposed algorithm under different missing probability conditions. The results show that the proposed algorithm outperforms the traditional non-cascading optimization algorithm, regarding clutter suppression and effectively improves the detection ability of radar.