Automated Testing of Spatially-Dependent Environmental Hypotheses through Active Transfer Learning

TL;DR

This work combines transfer learning and active learning through a Multi-Task Gaussian Process and an information-based objective function that can explore the space of hypothetical inter-quantity relationships and evaluate these hypotheses in real-time, allowing this new knowledge to be immediately exploited for future plans.

Abstract

The efficient collection of samples is an important factor in outdoor information gathering applications on account of high sampling costs such as time, energy, and potential destruction to the environment. Utilization of available a-priori data can be a powerful tool for increasing efficiency. However, the relationships of this data with the quantity of interest are often not known ahead of time, limiting the ability to leverage this knowledge for improved planning efficiency. To this end, this work combines transfer learning and active learning through a Multi-Task Gaussian Process and an information-based objective function. Through this combination it can explore the space of hypothetical inter-quantity relationships and evaluate these hypotheses in real-time, allowing this new knowledge to be immediately exploited for future plans. The performance of the proposed method is evaluated against synthetic data and is shown to evaluate multiple hypotheses correctly. Its effectiveness is also demonstrated on real datasets. The technique is able to identify and leverage hypotheses which show a medium or strong correlation to reduce prediction error by a factor of 1.4--3.4 within the first 7 samples, and poor hypotheses are quickly identified and rejected eventually having no adverse effect.

Figures (12)

• Figure 1: The ACFR's Swagbot autonomously collecting and analyzing soil samples.
• Figure 2: Example GP maps. Yellow and black indicate high and low values, respectively, and green dots show sample locations. GP produce smooth interpolations between measurements and higher uncertainty away from sample locations.
• Figure 3: Abstract depiction of the MTGP architecture. Gray circles represent values for each of the quantities, with $1$ labeling the QOI. The covariance function allows transfering information between quantities and representing how much they correlate (hypotheses).
• Figure 4: Maps of the QOI and prior quantities with varying levels of dependence. The top-right has high dependence with the QOI (H), the bottom-left has medium dependence (M), and the bottom-right has low dependence (L).
• Figure 5: Mean hypothesis scores between each quantity and the QOI throughout simulation runs of $30$ samples. Each subplot is a different combination of High, Medium, and Low hypotheses with values averaged over all runs.