Performance improvement via AR modeling based linear prediction for time division duplex (TDD) smart antenna systems

Abstract

In time-division-duplex (TDD) mode wireless communication, downlink performance of a smart antenna system (SAS) can be degraded due to variation of spatial signatures especially in vehicular scenarios. To mitigate this, prediction based downlink beamforming can be applied, which relies on using updated weight vectors via linear prediction of spatial signatures in the downlink interval based on their autoregressive modeling in the uplink interval. Here, we demonstrate the effectiveness of employing predicted spatial signatures as downlink weight vectors under varying mobile speed (V), multipath angle spread (/spl Delta//spl theta/), prediction filter order (P) and multipath number (L). We observe that in fixed Doppler shift conditions, i.e., Doppler shifts in the multipaths are integer multiple of Doppler shift in the dominant path, prediction based beamforming achieves better SNR improvements in the received signal at the mobile terminal with increasing V, P, and L. However its performance is not significantly affected by the variation in /spl Delta//spl theta/.