Online Informative Path Planning Using Sparse Gaussian Processes

Abstract

Estimating the environmental fields for large survey areas is a difficult task, primarily because of the field's spatio-temporal nature. A good approach in performing this task is to do adaptive sampling using robots. In such a scenario, robots have limited time to collect data before the field varies significantly. In this paper, we suggest an algorithm, AdaPP, to perform this task of data collection within a constraint on sampling time and provide an approximation of the environmental field. We test our performance against conventional sampling paths and show that we are able to obtain a good approximation of the field within the stipulated time.