A Decision Making Framework for Robot Companion Systems Usable by Non-Experts
In the course of becoming an increasingly important part of society, robots have also found their way into private households. For now, most robots are designed to solve only one or a few specific tasks. In the (near) future, however, robots are supposed to become companions assisting humans in their everyday life. A serious problem lies in the fact that requirements made upon robots as well as their fields of duty are largely dependent on the individual demands of the user. Due to this reason, the behaviour and the possible applications of a robot companion need to be customizable. The aim of this thesis is to develop a decision making framework for robot companions which offers solutions for the previously described challenges in the creation of robot companion systems. First of all, a suitable decision making algorithm that is applicable for variant tasks without a multitude of parameters having to be adjusted manually is created. This is important in order to give users without programming skills or technical expertise the possibility of enhancing the capabilities of their robot to a certain extent. The developed algorithm then is evaluated in a simulation in which human decisions are compared to decisions made by the algorithm. In addition to the evaluation made in a simulation, the decision making algorithm is implemented on the humanoid NAO robot. A modular software architecture is used in order to ensure that enhancements/modifications can be implemented without huge effort. Furthermore, the implementation provides interfaces making it possible to create new applications without programming by an XML configuration file. Based on these interfaces a tool assisting users without technical expertise in the creation of new applications is developed. Moreover, a usability study is conducted to reveal how the tool can be enhanced. Finally, the whole approach is evaluated via two human-robot interaction studies. Those studies aim at investigating how the participants perceive the robot’s decision making behaviour and if they can imagine using such a robot at home.