Automated Programming of an Industrial Robot through Teach-By Showing

Conventional programming methods require the user to learn the syntax and semantics of a proprietary robot programming language and to debug the program by iteratively editing, compiling, and executing the program. If the task to be performed by the robot requires sensor feedback (e.g., force or touch), the programmer must speculate on the proper sensor values and guardedly wait by the “emergency-stop” button in case of error. In this work, we investigate methods to automatically produce sensor-driven procedural programs from observations of successful physical demonstration by human teachers.