Situation Awareness for robotic platforms
Situation awareness is an important concept for the success of autonomous data acquisition campaigns performed with robotic boats. Our work focuses on the problem of detecting, modeling and interpreting robotic boat states with data-driven methods. By state we mean an abstract, compact and informative descriptor of key properties of the robot-environment system. In particular, we aim at developing interpretable models of boat states from traces of sensor data acquired during water-monitoring campaigns, by means of machine learning and artificial intelligence methods.
Our proposed solution focuses on unsupervised approaches, namely clustering and time series segmentation (e.g., SubCMedians, TICC and IHMM). These approaches are able to split multivariate time series into groups of observations corresponding to system states that have common properties and that can be compactly represented by mathematical models. Results show that the unsupervised methods we used can segment sensor data and identify key variables to determine interesting situations for the robotic boats (such as upstream or downstream navigation).
Authors: Alberto Castellini, Manuele Bicego Francesco Masillo, Maddalena Zuccotto, and Alessandro Farinelli