Central to robot exploration and mapping is the task of persistentlocalization in environmental fields characterized by spatially correlatedmeasurements. This paper presents a Gaussian process localization (GP-Localize)algorithm that, in contrast to existing works, can exploit the spatiallycorrelated field measurements taken during a robot's exploration (instead ofrelying on prior training data) for efficiently and scalably learning the GPobservation model online through our proposed novel online sparse GP. As aresult, GP-Localize is capable of achieving constant time and memory (i.e.,independent of the size of the data) per filtering step, which demonstrates thepractical feasibility of using GPs for persistent robot localization andautonomy. Empirical evaluation via simulated experiments with real-worlddatasets and a real robot experiment shows that GP-Localize outperformsexisting GP localization algorithms.
展开▼
