With the widespread application of smart antennas in 5G communication and radar detection, adaptive beamforming technology based on deep learning has become a research focus for improving the anti-interference performance of antenna arrays due to its powerful nonlinear modeling capability. It can transform the beamforming problem into a neural network regression problem, enabling the model to rapidly output an approximately optimal beamforming weight vector without prior information. Aiming at the issues of poor adaptability to dynamic interference and high computational complexity of traditional algorithms, this paper proposes IRDSNet, a novel adaptive beamforming algorithm based on Inception-ResNet-dual-pool Squeeze-and-Excitation Network (DP-SENet), to optimize the performance of uniform circular array antennas. IRDSNet integrates the Inception structure, depthwise separable convolution, and Ghost convolution to construct a multi-scale feature extraction module, enhancing the model’s feature extraction capabilities while maintaining a low parameter count. By introducing an improved DP-SENet, the model’s ability to focus on key features is enhanced, while the incorporation of residual modules optimizes feature transmission efficiency. Simulation results demonstrate that the IRDSNet algorithm achieves a null depth exceeding −90 dB at various interference angles, with an output Signal-to-Interference-plus-Noise Ratio (SINR) consistently above 23 dB and a short inference time, demonstrating excellent interference suppression performance.

This paper presents the effects of radio frequency interference (RFI) mitigation on a radio telescope’s sensitivity and beam pattern. It specifically explores the impact of subspace-projection mitigation on the phased array feed (PAF) beams of the Australian SKA Pathfinder (ASKAP) telescope. The goal is to demonstrate ASKAP’s ability to make science observations during active RFI mitigation. The target interfering signal is a self-generated clock signal from the digital receivers of ASKAP’s PAF. This signal is stationary, so we apply the mitigation projection to the beamformer weights at the beginning of the observation and hold them fixed. We suppressed the unwanted narrowband signal by 31 dB, to the noise floor of an 880 s integration on one antenna, with a typical degradation in sensitivity of just 1.5%. Sensitivity degradation over the whole 36 antenna array of 3.1% was then measured via interferometric assessment of system equivalent flux density (SEFD). These measurements are in line with theoretical calculation of noise increase using the correlation of the beam weights and RFI spatial signature. Further, degradation to the main beam’s gain is $\pm$ 0.4% on average at the half-power point, with no significant change to the gain in the first sidelobe and no variation during extended observations; also consistent with our modelling. In summary, we present the first demonstration of mitigation via spatial nulling with PAFs on a large aperture synthesis array telescope and assess impact on sensitivity and beam shape via SEFD and holography measurements. The mitigation introduces smaller changes to sensitivity than intrinsic sensitivity differences between beams, does not preclude high dynamic range imaging and, in continuum 1 MHz mode, recovers an otherwise corrupted holography beam map and usable astronomical source correlations in the RFI-affected channel.