AI-based cognitive radio based on Walsh signal processing for next generation terahertz wireless communication

In this thesis, we will utilize a predominantly end-to-end deep learning methodology, to enable the design of transceivers for 6G bands with more than 10 GHz of bandwidth. First, we want to make the first original digital sub-THz wireless communication system that applies the Walsh signal processing technique for sending and receiving messages adapted for high bandwidth scenarios. These models are further enhanced assuming imperfect channel knowledge. State-of-the-art auto-encoder models will have to be adapted to achieve higher constellations and even multicarrier communication in high bandwidth frequency selective channels. To enable channel estimation and equalization, we will explore semi-supervised learning methods with minimal training samples. Finally, the approach will be extended to work also in interference scenarios, to achieve truly cognitive THz radios for the 6th generation communication system.