Ultrawideband technology can be used to measure the distance between two device equipped with ultrawideband transceivers. Multiple ultrawideband devices known as anchors can localize a device that supports ultrawideband communication after each anchor measures the distance between itself and the device. The anchor and the tag can measure the distance between themselves by measuring the time of flight of the ultrawideband messages exchanged between them. If the location of the anchors are known, the location of the device can be computed. However, errors in range measurements can cause errors in localization of the device. There are many sources that can cause errors in the measured range between the anchor and the tag. For example, when the tag transmits a signal, and the path between the anchor and the tag is blocked, the signal may be attenuated to such an extent that it may go undetected at the anchor, causing the anchor to measure the time of arrival of a delayed reflection. In this thesis, we aim to find a solution to improve the ranging accuracy when an anchor is equipped with two antennas. We target approaches that can be implemented on an embedded processor with limited resources. In order to improve the ranging accuracy, we first investigate various machine learning based classifiers that can identify non-line of sight and line of sight signals. Next, for each of the classified signal, we train and test various machine learning based regression models that will predict the true distance from the distance measured by the anchor. We conclude the thesis, by assessing the robustness of the derived classification and regression models by evaluating the range improvements in measurements taken in different scenarios.