I am interested in autonomous robots that explore dangerous or inaccessible environments for the benefit of human knowledge. Understanding what features to visit, which instruments to deploy, and what data or samples to bring back are challenging perception issues. My current research tackles these problems with onboard machine learning. We leverage learning techniques to interpret data from multiple sensors, identify science targets and model the targets' distribution in the explored environment. This enables the exploration robot to identify scientifically-interesting trends and anomalies.
|