Achieving Robust Channel Estimation Neural Networks by Designed Training Data

Channel estimation is crucial in cognitive communications, as it enables intelligent spectrum sensing and adaptive transmission by providing accurate information about the channel state information. Current channel estimation neural networks are frequently tested by training and testing on one example channel or similar channels. However, data-driven methods often degrade on new data which they are not trained on, because they cannot extrapolate their training knowledge. This motivates us to investigate how to achieve neural network solutions that perform robustly over a wide range of realistic channels, but without any actual channel information being known at design time. In this paper, we propose design criteria to generate synthetic training datasets for neural networks, which guarantee that after training the resulting networks achieve a certain mean squared error (MSE) on a wide range of unseen channels. Therefore, these neural network implementations require no prior information of channels or parameters update for real-world deployment. Based on the proposed design criteria, we further propose a benchmark training dataset design which ensures intelligent operation for different channel profiles. To demonstrate general applicability, we use neural networks with different levels of complexity to demonstrate the generalization achieved. From simulations, neural networks achieve robust generalization to wireless channels with both fixed channel profiles and variable delay spreads.