Abstract

Sensor planning and active sensing, long studied inrobotics, adapt sensor parameters to maximize a utilityfunction while constraining resource expenditures. Herewe consider information gain as the utility function. Whilethese concepts are often used to reason about 3D sensors,these are usually treated as a predefined, black-box, com-ponent. In this paper we show how the same principles canbe used as part of the 3D sensor. We describe the relevant generative model forstructured-light 3D scanning and show how adaptivepattern selection can maximize information gain in anopen-loop-feedback manner. We then demonstrate howdifferent choices of relevant variable sets (correspondingto the subproblems of locatization and mapping) lead todifferent criteria for pattern selection and can be computedin an online fashion. We show results for both subproblems with several pattern dictionary choices and demonstrate their usefulness for pose estimation and depth acquisition.