Pilot Signal Design and Direct Ranging Methods for Radio Localization Using OFDM Systems

Having accurate localization capability is becoming important for existing and future terrestrial wireless communication systems, in particular for orthogonal frequency-division multiplexing (OFDM) systems, such as WiMAX, wireless local area network, long-term evolution (LTE) and its extension LTE-Advanced. To obtain accurate position estimates, not only advanced estimation algorithms are needed but also the transmitted signals should be scrutinized. In this dissertation, we investigate how to design OFDM pilot signals and propose and evaluate high accuracy ranging techniques with tractable computational complexity for localization.

We first employ an important tool from radar theory, the ambiguity function, to assess the accuracy of joint delay and Doppler shift estimation using a certain pilot signal. Accordingly, an optimal pilot signal should lead to an ambiguity function with a narrow main-lobe and low side-lobes. It is found that the equispaced and equipowered pilot signal (as used in LTE) results in an ambiguity function with high side-lobes. We propose to use the Cram\'{e}r-Rao bound in combination with the normalized side-lobe level (NSL) of the ambiguity function as figures of merit to devise the pilot signals. We then formulate the pilot signal design problem as a constrained optimization problem for which we propose a genetic algorithm to compute close-to-optimal solutions. The proposed method is a sound choice in a single-path scenario and a multi-path scenario with separable path components.

For scenarios where the number of path components is unknown and these components are not necessary separable, we propose a direct ranging technique using the received frequency-domain OFDM pilot signals. Compared to conventional (two-step) ranging methods, which estimate intermediate parameters such as the received signal strength, time-of-arrival, and biases introduced by non-line-of-sight (NLOS) propagation, etc., the direct ranging approach estimates the range from the received signal in one-step. This approach has the merit that it avoids LOS and first-path detection problems, the requirement of knowing the number of path components and it relaxes the separability condition of the path components. Employing a point process formulated channel model, which allows us to compute the necessary moments of the received signal, we propose and evaluate non-Bayesian and Bayesian range estimators. We show by means of Monte Carlo simulations that the proposed estimators, while exhibiting a tractable computational complexity, yield a significant ranging accuracy gain as compared to the non-coherent correlator-based estimator.