deletions | additions       diff --git a/abstract1.tex b/abstract1.tex  index e69de29..c829221 100644  --- a/abstract1.tex  +++ b/abstract1.tex 
...
Widespread possibilities make Micro Air Vehicles (MAVs) flexible  platforms. However, in confined spaces, specific localization  techniques are needed, which require high precision, reliability, and  rigorous error handling to minimize risks for people and hardware. Due  to their small size, MAVs have limited processing power and cannot  fall back to standard localization techniques. To address this  issue, this thesis describes an efficient onboard localization  technique for estimating $x,y$-coordinates for MAVs using a machine  learning-based approach, within a known, modifiable indoor  environment.  The computational power of the approach is scalable to  fit to different platforms.  The thesis presents the development, software and hardware  implementation, and results of a localization system that is based on  textons---small, characteristic image patches. The histograms of  textons are used as features for the $k$-Nearest Neighbors ($k$NN)  algorithm. The outputs of this regression technique--multiple possible  $x, y$-coordinates are used in a particle filter to neatly aggregate  and smooth the estimates. The estimates of the computer vision-based  system are compared to the ground truth in five on-ground and two  in-flight experiments. Promising results were obtained for all tasks:  waypoint navigation, accurate landing, and stable hovering in the  indoor environment.  To predict the performance of the proposed algorithm in different environments, an evaluation  technique is developed that compares actual histogram similarities to  ideal histogram similarities based on distance between the underlying  positions. The technique assigns a loss value to a given set of  images, allowing to compare different environments. Therefore, it  allows for spotting difficult or ambiguous locations or safe landing  spots. Additionally, a tool is presented that creates synthetic images  that could be taken during an actual flight. The synthetic images are  used to compare 46 possible maps---images with a high resolution. The  best one, the worst one and the one with median loss were printed out  to compare their performance in the real-world. In fact, the results  could be replicated on these maps.  The approach is based on three pillars: (i) a shift of processing power to an pre-flight phase to pre-compute  computationally complex steps, (ii) lightweight and adaptable algorithms to ensure real-time performance and portability to different platforms,  (iii) modifiable environments that can be tailored to the proposed algorithm.