One of the main challenges to designing robots that can operate around humans is to create systems that can guarantee safety and effectiveness, while being robust to the nuisances of unstructured environments, from hardware faults to software issues, erroneous calibration, and less predictable anomalies, such as tampering and sabotage. However, the fact that the streams of observations and commands possess coherence properties suggests that many of these disturbances could be detected and automatically mitigated with general methods that imply very low design efforts. Currently, robotic systems are developed as a set of components realizing a directed "computation graph". This project focuses on theoretical methods, applicable designs, and reference implementation of a faults/anomalies detection mechanism for low-level robotic sensorimotor signals. The system, without any prior information about the robot configuration, should learn a model of the robot and the environment by passive observations of the signals exposed in the computation graph, and, based on this model, instantiate faults/anomalies detection components in an augmented computation graph.
The project engages undergraduate and graduate students in advanced robotics design and development. It is expected the research results will have a significant impact on future robotic systems and machine learning.
