Intuitive Robot Programming for Novice and Expert Users

Robots have revolutionized industrial automation. Automation has many advantages---for employees and companies alike---and counteracts the shortage of skilled workers. That is why many companies are striving to increase the level of automation. However, one obstacle is the programming effort required to adapt a robotic system to new or changed tasks. The reason for this is that many robots are still programmed in the classic textual way, which requires experts and time. Therefore, to advance automation, the programming of robots must become easier. In this thesis, concepts are developed and evaluated that increase the usability of robot programming. These concepts are based on Task-Level Programming (TLP), in which experts develop generic robot capabilities (skills) that nonexperts can instantiate and parametrize. The developed concepts facilitate TLP for both sides. For experts, a programming concept based on hierarchical state machines is developed that facilitates the development of skills with a graphical programming environment. Several application examples are used to show how autonomous behavior can be achieved even with complex systems. For laypeople, a graphical user interface is presented, with which skills can be intuitively sequenced and parametrized. By means of geometric primitives, recurring tasks can be programmed in a time-efficient manner. Two user studies demonstrate the usability of the software. Another concept allows the user to operate the interface via arbitrary input devices. To bring the expert and layperson sides together, an interface component is developed. The component serves as a link to make skills available to laypeople and to let them be executed and monitored. A major advantage of this component is the possibility to define parametrization processes, which simplifies the parametrization of skills. Finally, based on the above concepts, a new programming method is developed that combines the advantages of Programming by Demonstration and TLP. A user study shows how this procedure is many times faster and has a higher usability compared to pure TLP