Automated Programming and Skill Transfer 

Abstract

One of the key enablers of smart work-cells where sophisticated assembly tasks are to be carried out by robots under a poorly structured environment is how to automate robot programming with a minimal human effort.

This talk addresses an approach to robot self-programming based on learning from the observation of human assembly workers and automatically transforms whatever learned into executable robot programs.

More specifically, the proposed approach involves in deep-learning based understanding of the intentions of assembly workers embedded in the sequence of their individual assembly operations and by translating their intentions identified automatically into a sequence of executable robot actions and programs, if necessary, with the assistance of assembly specifications. A prototype of the proposed automated robot programming system has been built for the assembly of electronic equipment to demonstrate the feasibility of the proposed approach. The results show that the proposed automated robot programming is not only feasible but also viable with a dramatic simplification of human effort in completing robotic work-cell programming under a saving of a great amount of time and energy.