This paper presents the frontier tree exploration algorithm, a novel approach to autonomously explore unknown and unstructured areas. Focus of this work is the exploration of domestic environments with lots of arbitrary obstacles, for example furbished appartements. Existing and well-studied approaches like greedy algorithms perform worse when obstacles are included and the range of distance sensors is limited. Base of this research is the frontier tree. This data structure offers two main features. Firstly it serves as a memory of past poses during exploration, where boundaries between known and unknown space are inserted into the tree. Secondly, the data structure is then utilized to decide between future navigation goals. This approach is compared to the class of nearest neighbor exploration algorithms and a decision theoretic approach. The algorithm is tested in a simulation study with furbished ground maps and in a real life domestic environment. The paper shows, that frontier trees exhibit a superior performance of distance traveled and the number of exploration steps compared to a nearest neighbor algorithm.