Local Subdivisional Channel Estimation in RIS-Assisted THz MIMO-OFDM Systems

Reconfigurable intelligent surfaces (RIS) offer significant performance gains for terahertz (THz) multiple-input and multiple-output (MIMO)-orthogonal frequency division multiplexing (OFDM) systems. However, channel estimation (CE) in high-dimensional RIS channels remains challenging. This paper proposes a compressed sensing (CS)-based CE method that separates angle and gain estimation, reducing pilot overhead and computational complexity. We leverage spatial path selection enabled by designed beam and reflection patterns to reduce dimensionality. Our method achieves over $35 \%$ pilot overhead reduction and significantly improves computational efficiency, while maintaining accuracy comparable to existing methods.