Online Learning Techniques for Improving Robot Navigation in Unfamiliar Domains

Abstract

Many mobile robot applications require robots to act safely and intelligently in complex unfamiliar<br>environments with little structure and limited or unavailable human supervision. As a<br>robot is forced to operate in an environment that it was not engineered or trained for, various aspects<br>of its performance will inevitably degrade. Roboticists equip robots with powerful sensors<br>and data sources to deal with uncertainty, only to discover that the robots are able to make only<br>minimal use of this data and still find themselves in trouble. Similarly, roboticists develop and<br>train their robots in representative areas, only to discover that they encounter new situations that<br>are not in their experience base. Small problems resulting in mildly sub-optimal performance are<br>often tolerable, but major failures resulting in vehicle loss or compromised human safety are not.<br>This thesis presents a series of online algorithms to enable a mobile robot to better deal with<br>uncertainty in unfamiliar domains in order to improve its navigational abilities, better utilize<br>available data and resources and reduce risk to the vehicle. We validate these algorithms through<br>extensive testing onboard large mobile robot systems and argue how such approaches can increase<br>the reliability and robustness of mobile robots, bringing them closer to the capabilities<br>required for many real-world applications.