Authors:
Pinky Thakkar
;
Steven M. Crunk
;
Marian Hofer
;
Gabriel Cadden
;
Shikha Naik
and
Kim T. Ninh
Affiliation:
San Jose State University, United States
Keyword(s):
Adaptive Sampling, Bayesian Inference, BRDF, Maximum Entropy, Optimal Location Selection.
Related
Ontology
Subjects/Areas/Topics:
Autonomous Agents
;
Informatics in Control, Automation and Robotics
;
Robotics and Automation
;
Vehicle Control Applications
Abstract:
Traditional methods of data collection are often expensive and time consuming. We propose a novel data collection technique, called Bayesian Adaptive Sampling (BAS), which enables us to capture maximum information from minimal sample size. In this technique, the information available at any given point is used to direct future data collection from locations that are likely to provide the most useful observations in terms of gaining the most accuracy in the estimation of quantities of interest. We apply this approach to the problem of estimating the amount of carbon sequestered by trees. Data may be collected by an autonomous helicopter with onboard instrumentation and computing capability, which after taking measurements, would then analyze the currently available data and determine the next best informative location at which a measurement should be taken. We quantify the errors in estimation, and work towards achieving maximal information from minimal sample sizes. We conclude by pr
esenting experimental results that suggest our approach towards biomass estimation is more accurate and efficient as compared to random sampling.
(More)