Deep Homography Estimation-Based Heterogeneous Decoupled Channel Precoding Mapping

Abstract

The fully decoupled radio access network (FD-RAN) is a potential 6G RAN architecture offering flexibility and efficiency. However, its decoupling nature poses significant challenges for real-time downlink (DL) channel precoding. To overcome the diminishing uplink(UL)-to-DL correlation in FD-RAN, this letter introduces a deep homography estimation-based heterogeneous decoupled channel precoding mapping. Inspired by the analogy between the mapping of heterogeneous UL and DL channels from decoupled base stations (BSs) to user equipment (UE) and the homography between two planes, the method utilizes UL channels to achieve both DL precoding and UE localization. The approach demonstrates superiority within a typical FD-RAN scenario under two challenging heterogeneous decoupled channel setups and various training overheads.