This research proposes innovative AI solutions to overcome three key challenges in terahertz ultra-massive MIMO systems: "hard to compute," "hard to model," and "hard to measure."

Key Insights:

• Presents three research roadmaps for AI integration in THz UM-MIMO
• Advocates for model-driven deep learning that leverages domain knowledge
• Explores potential of foundation models for wireless communications
• Addresses engineering challenges in next-generation wireless networks

This work has significant implications for the engineering of 6G networks and beyond, laying groundwork for AI-enabled wireless systems that can handle the extreme complexity of terahertz communications with unprecedented efficiency and accuracy.

AI and Deep Learning for THz Ultra-Massive MIMO: From Model-Driven Approaches to Foundation Models