Adaptive detection of complex MIMO signals in mines based on transfer learning

To address reduced robustness and accuracy in signal detection models for wireless channels in complex mining environments during online deployment, we explored the Conditional Adaptive Detection Network (CADN). Leveraging the Multi-Domain Maximum Mean Discrepancy (MMD) method to enhance pre-trained model robustness, the detection network is adapted to dynamically changing and complex mining environments with channel fading. Incorporating the multi-modal characteristics of MIMO signals, the model dynamically adapts using Conditional Domain Adversarial Networks (CDAN) for adaptive signal detection. Experimental results show significant improvements in both robustness and accuracy in dynamically changing mining environments with Nakagami fading MIMO systems.