Abstract
This paper proposes an over-the-air (OTA) channel emulation methodology within the multi-probe anechoic chamber (MPAC) framework to test artificial intelligence (AI)-based spatial-domain beam management (BM) for millimeter wave devices. The reconstructed channel provides reliable measurements as AI input and accurate ground truth for validating AI predictions in instantaneous multi-beam environments. Base station (BS) filtering is applied to the Cluster Delay Line (CDL) channel model, which results in rapid fluctuations in the propagation environment at the user equipment (UE) side during the BM procedure. Two novel approaches-approach A and approach B-are proposed to emulate the instantaneous multi-beam channel environment, with approach A suitable for emulating random target beams and approach B for pre-defined ones, respectively. Both approaches are based on the Beam Power-Angular-Spectrum Similarity Percentage (BPSP), a new metric introduced to evaluate the channel accuracy of reconstruction. Simulations and experimental results are presented to validate the accuracy of the channel reconstruction. Furthermore, the performance of an actual AI model deployed at the UE side is evaluated using the proposed methodology.
Funding Information
  • National Key Research and Development Program of China (2023YFF0717804)
  • Fundamental Research Funds for the Central Universities (2242022k60008)

This publication has 30 references indexed in Scilit: