Gaussian Process Regression for Learning Environment Impacts on Localization Accuracy of a UAV with Respect to UGV for Search Planning

In this article, we present a path planning algorithm for a team of an Unmanned Ground Vehicle and an Unmanned Aerial Vehicle (UAV) that leverages Gaussian process regression to plan a path that meets information gathering objectives while reducing the UAV's localization uncertainty by learning to compensate for outlier measurements or missed expected sensor measurements over the trajectory. Simulation results are compared to approach that also compensates for belief space planning but is incapable of handling outliers or unexpected degradation from the environment1.