Examenspresentation av Joakim Eeby & Thiago Puppin Romano (MPICT)

Supervisors: Xinlin Zhang, Anders Aronsson (Ericsson Research); Hao Guo, Mehdi Sattari; Tommy Svensson (Chalmers)
Examiner: Tommy Svensson (Chalmers)

Abstract:
Accurate channel state information (CSI) is a key component for efficient and reliable communications in wireless networks. Utilizing CSI enables the system to adapt to various channel conditions and increases the total throughput. In 5G New Radio (NR), codebooks determine how the user equipment (UE) reports CSI to the gNodeB (gNB). More specifically, a precoder matrix indicator (PMI) is included in the CSI report, suggesting the recommended precoder matrix that the gNB may use for downlink transmission. Since the channel is time-varying, the CSI report may be outdated when received at the gNB, which degrades system performance. A potential solution is to predict the CSI at the gNB.

Thus, this study investigates CSI prediction based on the current codebooks in 5G NR. Moreover, a new reporting format is proposed where the throughput is increased with only a minor increase in overhead. An autoregressive (AR) model combined with a Kalman filter is used for prediction. The results indicate that prediction is possible using existing codebooks, but by slightly increasing the overhead using the new reporting format, the channel can be recreated at the gNB and more accurately predicted. Simulations based on a standardized channel model showed that implementing a Kalman filter on existing codebooks provided a gain of around 1.9 dB. By using the new reporting method, a gain of approximately 2.2 dB was obtained.

Welcome!
Joakim, Thiago and Tommy