Goal-Oriented End-User Development of Robots

End-user development (EUD) tools must balance a user's desired level of control over a robot's task with the robot's ability to plan and act autonomously. Many existing task-oriented EUD tools enforce a specific level of control, e.g., by requiring that users hand-craft detailed sequences of actions. There is a lack of EUD solutions that offer users the flexibility to choose the level of task detail they wish to express. We thereby created a novel EUD system, Polaris, which affords users a similar level of expressiveness as state-of-the-art EUD systems, yet the fundamental building blocks of tasks within Polaris are goals rather than robot actions. Polaris's goal-oriented programming paradigm enables users to express high-level task objectives or lower-level task checkpoints at their choosing, while an off-the-shelf task planner fills in any remaining task detail. To ensure that goal-specified tasks adhere to user expectations of robot behavior, Polaris is equipped with a Plan Visualizer that exposes the full task plan to the user prior to runtime. In what follows, we describe the system design of Polaris and its evaluation with 32 human participants. Our results support the Plan Visualizer's ability to help users craft higher-quality task plans. Furthermore, there are strong associations between Plan Visualizer usage and perceptions of robot performance, and evidence that users' past familiarity with robots has a key role in shaping user experience.