Efficient techniques for OFDM applications over mobile radio channels

Abstract

This Thesis is oriented to the study and proposal of modeling solutions, as well as robust techniques, for low computational complexity channel estimation and prediction for OFDM systems operating over mobile wireless channels. It is also oriented to the development of resource allocation techniques for multiple access systems in this context, which include the typical imperfections of practical implementations of the physical layer in the problem formulation. The main contributions can be grouped as follows.

Parametric and statistical models commonly used for wireless channels are studied, and a novel hybrid modeling technique is developed. This novel technique uses statistical channel information related to the type of mobile applications considered, together with a parametric structure based on the decomposition of the channel trajectory over time into a known orthonormal basis. This channel model allows a simple evaluation of the number of parameters to be estimated, and also allows a trade-off between modeling error and the complexity of the associated structure.

Based on the formulated hybrid channel model, and the distinctive parallelization characteristic of OFDM over a frequency selective fading channel, a time domain recursive channel estimator is developed capable of efficient tracking of the temporal evolution of a flat fading channel. Besides of leading to an estimator robust to the shape of the channel Doppler spectrum, the proposed formulation results in a low computational complexity estimator because direct estimation of the channel Doppler components is avoided, task which involves the use of highly elaborated spectral estimation techniques. Applying analogous concepts to the ones used to model the time variation of the channel, an structure is derived to extend the results to the case of frequency selective fading channels, which inherits the computational efficiency of the flat fading channel estimator as it is based on the same design criteria.

Fair distribution of channel resources on multiuser systems is a natural concern in multicarrier systems, being its implementation critical for the case of mobile channels, where channel state information varies significantly over resource allocation intervals. For this setting, an extension of the proposed estimation scheme is developed for application in the prediction of future channel state information. The developed solution results of lower complexity than similar solutions available in the literature, obtaining a sufficiently large prediction horizon for its application on current resource allocation schemes.

Finally, the performance of the proposed channel estimator/predictor structure is evaluated on a realistic multiuser environment and a characterization technique is proposed for the prediction error associated to practical channel predictors. The inclusion of this characterization in the system resource scheduler formulation, allows a significant improvement on its efficiency as it takes into consideration imperfect channel state information, unavoidable when considering practical implementations.