Long-term 'Localization of Mobile Robots in Dynamic Changing Environments

Long-term localization in dynamic changing environments is still a challenge in robotics. Traditional localization algorithms typically assume that the environment is static. However, in many real-world applications, such as parking lots and industrial plants, there are always dynamic objects (e.g. moving people) and semi-dynamic objects (e.g. parked cars and placed goods). In this paper we address this challenge by introducing a long-term localization algorithm in the environments which combine dynamic objects and semi-dynamic objects. Localizability-based-updating particle filter (LU-PF) algorithm is proposed here Not only we use localizability matric to build an updating mechanism, but also it is used for localization system Besides. we propose the dynamic factor as long-memory information to serve as prior knowledge, which improves the robustness of updating process. Experiments in parking lots demonstrate that our approach has better localization results with a more accurate up-to-date map compared to other methods.