Reliable long-term localization is the essential requirement for home mobile robots to work in realistic applications. Despite the subsequent evolution in mobile robot localization and SLAM during last decades, in the ever changing home environments, there still exist challenges to be solved before mobile robots can reliably localize themselves during long periods of time.

In this thesis, a reliable long-term localization solution for home mobile robots is developed. We start by doing SLAM to get the map of the environment. We proposed a computationally efficient SLAM approach that is robust in dynamic environment based on real-time sliding window temporary map consistency test. Then, normal localization is executed. A variation of the branch-and-bound correlative matching algori...[ Read more ]

Reliable long-term localization is the essential requirement for home mobile robots to work in realistic applications. Despite the subsequent evolution in mobile robot localization and SLAM during last decades, in the ever changing home environments, there still exist challenges to be solved before mobile robots can reliably localize themselves during long periods of time.

In this thesis, a reliable long-term localization solution for home mobile robots is developed. We start by doing SLAM to get the map of the environment. We proposed a computationally efficient SLAM approach that is robust in dynamic environment based on real-time sliding window temporary map consistency test. Then, normal localization is executed. A variation of the branch-and-bound correlative matching algorithm is presented, which can provide fast and guaranteed optimal global localization result in home environment. Finally, to update the environment change information, the localization map is updated via temporary SLAM. In order to avoid the damage of the localization map due to map update, we present a real-time map consistency test algorithm that can test temporary map quality. The robustness of the this long term localization approach are validated via a number of real world experiments in a variety of home environments.