Logo TU Ilmenau




Localization in Wireless Networks: Improvement of Localization Techniques

Dr.-Ing. Oleksandr Artemenko
Date of publication
Wireless communications have already become a very important part of our everyday life. According to this, hundreds of new applications emerge every week defining higher and higher requirements on hardware and software. Localization is one of the most crucial issues for many such applications. To keep it simple and cheap, but nevertheless accurate and robust, is still a big challenge for thousands of researchers all over the world.

Disasters represent one of the special scenarios with high requirements on localization results. Uncertainties in the working environment, only few data available, and tight time constraints are only some problems that emerge thereunder. Much scientific research has been conducted in the area of localization techniques. However, none of them may be applied to our disaster scenario with no or little modification. Are there any other ways leading to a solution?

Development of new, more robust technologies represents one possible way to solve this problem and satisfy the needs in precise localization. Another way is to improve existing technologies while providing new and more sophisticated schemes that replace existing ones or extend them enabling even more accurate results. We propose in this thesis to follow the second solution. Using only standard hardware, which is available on most of platforms, we present in our work different refinement strategies that increase the accuracy and robustness of the "cheap" localization solutions.

First, this thesis divides the location estimation process into five main steps presenting its new and clear structure: parameter estimation, pre-improvement, distance estimation, position calculation and post-improvement. Here, the steps - parameter estimation, distance estimation, position calculation - represent the core of the localization process. The pre- and post-improvement steps deal only with the refinement of results and enable better location estimation accuracy. This work focuses primarily on the pre- and post-improvement stages, giving an overview about existing methods in these areas. Thereafter, new possibilities to improve the localization results are being developed, implemented, simulated and evaluated on the real testbed. To show the impact on the resulting localization accuracy, the core steps of the localization process will be applied every time while investigating different improvement strategies.

According to the pre-improvement step, we present different improvement strategies in selecting the most efficient constellation of reference information out of the redundant data available for localization. The developed methods, called Optimum Anchor Selection algorIthmS (OASIS), are being evaluated on the real data sets that were collected in our outdoor experiment using an unmanned aerial vehicle (UAV). The obtained results show significant improvement which can be achieved with OASIS schemes.

In the post-improvement step, this thesis proposes to use additional information about the network available during the localization. One concrete solution can be represented by distances between couples of mobile nodes being localized. Based on this information, we introduce a novel improvement technique called Universal Improvement Scheme (UnIS). A corresponding mathematical model is being developed and simulated on the appropriate simulation platform. Additionally, outcome from the simulations is being validated by the empirical results obtained in a deployed wireless sensor network. The results from both simulations and experiments indicate that the average localization error can be improved significantly using UnIS approach reaching an improvement ratio of 73%. Additionally, UnIS refinement is being compared to a well-known Kalman Filter (KF) technique. The results show that UnIS outperforms KF, presenting a much higher improvement ratio: 73% against 43%.
External link