Book Description
"In the first work a recurrent neural network (RNN) is employed for MIMO channel prediction. A novel PSO-EA-DEPSO off-line training algorithm is presented and is shown to outperform PSO, PSO-EA, and DEPSO. This predictor is shown to be robust to varying channel scenarios. New expressions for the received SNR, array gain, average probability of error, and diversity gain are derived. Next, a new expression for the outage capacity of a MIMO system with no CSI at the transmitter and an estimate at the receiver is presented. Since the outage capacity is a function of the first and second moments of the mutual information, new closed form approximations are derived at low and high effective SNR. Also at low effective SNR a new result for the outage capacity is presented. Finally, the outage capacity for a frequency selective channel is derived. This is followed by a MIMO RNN predictor that operates online. A single RNN is constructed to predict all of the MIMO sub-channels instantaneously. The extended Kalman filter (EKF) and real-time recurrent learning (RTRL) algorithms are applied to compare the MSE of the prediction error. A new expression for the channel estimation error of a continuously varying MIMO channel is derived next. The optimal amount of time to send training pilots is investigated for different channel scenarios. Special cases of the new expression for the channel estimation error lead to previously established results. The last work investigates the performance of a MIMO aeronautical system in a two- ray ground reflection scenario. The ergodic capacity is analyzed when the altitude, horizontal displacement, antenna separation, and aircraft velocity are varied"--Abstract, leaf iv.
