Thesis (M.S., Computer Science)--University of Idaho, June 2014 | The ability to easily train, and interact with robotic agents has been the focus of a great deal of research within the fields of Artificial Intelligence and Machine Learning. Learning from Demonstration (LfD) is a common approach for teaching robotic agents new behaviors because it requires little specialized knowledge of robotic hardware or of computer programming. Evolutionary learning techniques have also commonly been applied to robotic learning tasks, but the two techniques have rarely been used together due to the computational requirements of running an evolutionary process on-line, and on-board a robot during the LfD process. This thesis presents an on-line, on-board, evolutionary learning from demonstration protocol that enables autonomous robotic agents to learn how to complete a path following task. In addition, the robots are capable of recording their GPS location during autonomous operation which has a wide variety of practical applications from autonomous trail mapping, to marking objects of interest in disaster response applications.