Analysis of RSRP Prediction in Millimeter Wave Systems

In 5G millimeter wave (mmWave) systems, user equipments monitor the quality of beam pair links by measuring the reference signal received power (RSRP). This paper considers a proactive beam management design enabled by predicting future RSRP based on the past channel measurements. Based on a generic geometric channel model, we compute the RSRP prediction error of a linear MMSE predictor, and relate the prediction error to key system parameters, such as maximum Doppler spread and the periodicity of channel measurement. Our analysis reveals that only when the channel measurement periodicity is below a critical threshold, which is defined by the maximum Doppler spread frequency in the link, zero error prediction can be achieved asymptotically in the limit of the number of measurements. We further propose a beam selection approach based on the RSRP prediction to minimize the beam failure probability in the serving link. Numerical results show that the proposed approach provides a near optimal performance in terms of reducing the time duration in beam failure, comparing with a genie aided beam selection approach.