Link to System (L2S) Interfacing for Advanced Receiver Strategies

Abstract

The size and complexity of wireless communication networks have grown tremendously over the last few decades. Analysis of a wireless communication system requires computer simulations of the entire communication network, spanning multiple cells with a large number of base stations and mobile terminals. This normally involves complex physical layer computations in order to evaluate the receiver performance with the transmitted signals subjected to interference, multipath propagation, and shadowing. Link to system (L2S) interfacing reduces the computational complexity associated with the physical layer performance evaluation of multiple communication links by predicting the receiver behavior under different channel conditions using precalculated lookup tables (LUTs). This thesis investigates the L2S interfacing for different advanced receiver strategies using various nonlinear mapping functions. Different transmission scenarios such as single input single output, single input multiple output, and multiple input multiple output are considered. Besides using the conventional AWGN channel performance as the reference LUT, the mean of different channel frame error performance is also suggested as reference and the prediction accuracy of both have been compared. L2S framework has been implemented using the post detection signal to noise ratio (SNR) values as the received signal quality measure. The existing L2S work for SISO, linear MIMO systems has been extended to iterative and maximum likelihood receivers, where finding an accurate estimate of the received signal quality which is highly correlated to the receiver output is an open problem and needs to be fully explored. Algorithms for the post detection SNR value estimation for iterative and maximum likelihood receivers have been proposed and their prediction performance is validated for diverse communication channels. It is shown that, the post detection SNR value is an accurate measure of the quality of the received signal. However, for MIMO system with single stream encoding, the accurate estimation of the post detection SNR value for each individual link is not essential, but rather an accurate average value over multiple links is found to be sufficient. The other main contribution of this thesis is the formulation of a Artificial Neural Network (ANN) framework for the receiver performance prediction. ANN has been applied extensively to diverse applications due to its fast processing in real time scenarios. This is due to its ability to learn different tasks and to make decisions without being explicitly programmed. ANN has been used in this thesis for L2S interfacing in order to reduce the extensive training required for generating the reference curves in the classical L2S interfacing without having to compromise the quality of the prediction process. It has been shown that the quality features extracted from the received signal can be used in the machine learning algorithms for accurate prediction of the link level performances. An ANN based L2S interfacing has been implemented for different linear and nonlinear MIMO receiver strategies. The ANN based technique gives good link level performance prediction accuracy while doing away with the need for computing the pre-estimated LUTs required in classical L2S interfacing. Additionally, the ANN model when imported into the communication system simulation chain gives an independent frame error decision for each transmitted frame.