Predictive wireless antenna selection for high mobility conditions

Accurate channel knowledge is indispensable to the practical success of channel-aware wireless communication technologies. However, channel estimates obtained from pilot symbols rapidly become outdated due to fast time variations of multipath fading channels. To reduce system cost, antenna subset selection reduces radio frequency (RF) chain components. For systems selecting a subset of a plurality of antennas for reception, this outdated channel information is a significant impediment to selection and data decoding reliability. In this thesis, training-based schemes for antenna selection (AS) for time-varying channels which account for practical constraints such as training, packetization and antenna switching time are proposed based on discrete prolate spheroidal sequences. They only operate with knowledge or estimates of the Doppler frequency and the channel signal-to-noise ratio (SNR), but do not require detailed statistical correlation knowledge.;The second part of this thesis investigates the more general case of selecting a subset of a plurality of receive antennas. A new pilot-based receive antenna subset selection algorithm for time-varying frequency-flat channels is presented. The proposed AS algorithm is shown to outperform AS methods based on Fourier prediction/estimation as well as SIMO systems with perfect channel knowledge. Analysis of MPSK and quadrature amplitude modulation (MQAM) SEP for systems with receive AS is provided.;The combination of AS and orthogonal frequency division multiplexing (OFDM) over the more realistic time-varying and frequency-selective fading scenario is examined in the final part. Training schemes for receive AS using packet-rate antenna switching for SIMO and multiple input-multiple output (MIMO) OFDM systems are developed. The suggested schemes exhibit a superior performance over AS methods using either linear interpolation/extrapolation or Fourier prediction/estimation techniques.;A pilot-based AS scheme for time-varying frequency-flat channels for single input-multiple output (SIMO) systems selecting one of a plurality of antennas using packet or symbol-rate antenna switching is first proposed. It is demonstrated that the presented scheme provides significant performance gain over AS methods using Fourier-based orthogonal training as well as over single antenna systems with perfect channel knowledge. Analytical expressions for the symbol error probability (SEP) of M-ary phase-shift keying (MPSK) for systems employing the suggested techniques are provided.