AI Receiver Design and Power Allocation for Superimposed DMRS and Data in MIMO-OFDM

A recent paper on arXiv (2608.13809) introduces an AI-driven receiver aimed at superimposed demodulation-reference-symbol (DMRS) and data transmission in MIMO-OFDM frameworks. The researchers establish an analytical model to represent the iterative interaction between channel estimation (CE) and MIMO detection (MD) in an iterative CE and detection (ICED) process, which is utilized to enhance power distribution and pilot patterns. Additionally, they create an AI receiver based on a Transformer encoder that integrates the ICED framework. Simulation results indicate that the AI-ICED receiver utilizing SI-DMRS significantly boosts spectral efficiency when compared to traditional systems.