Global Visual And Semantic Observations for Outdoor Robot Localization

Most approaches to robot visual localization rely on local, global visual or semantic information as observation. In this paper, we use the combination of global visual and semantic information as landmark in the observation model of Bayesian filters. Introducing the improved Gaussian Process into observation models with visual information, we extend the GP-Localize algorithm to high dimensional data, which in this way the Bayesian filters can consider all the historical data with spatiotemporal correlation to achieve constant time and memory for persistent outdoor robot localization. Our another contribution is to combine the above parts into a system for robot visual localization and apply it to two real-world outdoor datasets including unmanned ground vehicle (UGV) and unmanned aerial vehicle (UAV). Through the analysis of experimental results, the features using global vision and semantic results will be more accurate than using single features. By using the combined features and improved Gaussian process approximation method in Bayes filters, our system is more robust and practical than existing localization systems such as ORB-SLAM.