Evaluation of Gaussian Processes for Large Scale Terrain Modeling
Hugh F Durrant-Whyte, Eric W. Nettleton, Fabio Tozeto Ramos, Shrihari Vasudevan · 2009
This paper addresses the problem of large scale terrain modeling for a mobile robot. Building a model of large scale terrain that can adequately handle uncertainty and incompleteness in a statistically sound way is a challenging problem. A recent work [Vasudevan et al., 2009] proposed non-stationary Gaussian processes (GP’s) based on the neural network kernel as a solution to the problem. GP’s naturally provide a multi-resolution representation of space, incorporate and handle uncertainty aptly and cope with incompleteness of sensory information. GP regression techniques may be applied to estimate and interpolate (to ll gaps in occluded areas) elevation information across the eld. This paper presents results evaluating GP’s for the problem of large scale terrain modeling. Extensive cross validation experiments were conducted on real world datasets obtained from dierent mine sites. These experiments are used to report statistically representative results that characterize the performance of the GP method for large scale and complex terrain modeling. They also compare its performance with grid based representations using many different interpolation techniques as well as triangulated irregular networks (TIN’s); these represent the state-of-the-art in large scale terrain modeling.