A Joint Optimization Method of Pulse Transmission and Range-Doppler Reconstruction for Pulse-Doppler Radar

To obtain a large unambiguous sensing range, traditional single-channel pulse-Doppler (PD) radar requires a large pulse repetition interval between adjacent pulses. However, due to the Nyquist sampling theorem, the pulse repetition frequency must be larger than the range of the Doppler domain, otherwise, aliasing in the Doppler domain will affect the range-Doppler (RD) reconstruction quality. Existing methods have designed non-uniform pulse transmission patterns to realize unambiguous range extension or develop compressive sensing-based algorithms to obtain high-quality RD reconstruction. However, the design of existing pulse transmission patterns and the RD reconstruction algorithms are often separated, which degrades the aliasing removal capability of non-uniform transmission. To solve this issue, this article introduces a data-driven joint optimization method that couples the learnable pulse transmission pattern and RD reconstruction algorithm, with the final RD reconstruction quality taken as the joint optimization objective. Specifically, we propose a learnable pulse transmission pattern and optimize the pulse transmission intervals in a random parameter freeze way. In addition, we propose a deep unfolding network for high-quality RD reconstruction based on approximate observations named MF-ISTA-Net. Experiments verify that the proposed method achieves superior RD reconstruction quality with the maintaining unambiguous sensing range extension in comparison with existing methods.