Overcoming Robotic Failures through Autonomicity

NASA is pursing ever increasing autonomous systems to perform new science and exploration. These missions will be out of contact with mission control for extended periods of time in very harsh and unforgiving environments. To successfully perform these missions autonomously, they will have to have autonomic properties. There have been recent studies of failures of robotic systems that examine how robots are failing in the field. By studying the types of failures that are occurring in robots autonomic properties can be developed for not only robots but for NASA autonomous missions to make them more robust. This paper examines the failures that were identified in a study of robot failures and maps them on to some autonomic properties that would address these failures. By classifying the failures and then mapping them to autonomic properties, classes of autonomic properties can be developed that can then be specialized for each particular mission