Robotic application development is challenging due to varying abstraction levels and domain concerns. While component-based architectures aid integration, programming and configuring these components requires deep domain knowledge. Current novice interfaces hide low-level details for accessibility but obscure the relationship between behavior and system details.
In the course of his dissertation, Thomas Witte resolves the conflict between abstraction and transparency by tagging and tracking data flow at runtime, gathering provenance information to relate data to higher abstraction levels. Source Location Tracking (SLT) uses this information to automatically apply user changes to corresponding inputs or data, ensuring accurate output after reexecution.
Key contributions include extending SLT within the Robot Operating System (ROS), analyzing ROS applications for consistency, and creating a novice programming environment for quadcopter missions. This environment allows direct manipulation of source code through live visualization. A user study with 82 participants confirmed the positive impact of SLT-enabled features on program correctness and learning success.