To date, drawings still serve as the main mean for assembly guidance. However, the use of drawing guidelines is not enough for the complexity of assembling large-scale apparatus such as computer systems or other intricate components. The challenge involves: (a) defining standard formats to describe the assembly process as well as defining a common storage method and a repository for assembly documentation to which refer to when needed; (b) creating and storing a well-documented self-documented accessible assembly guideline; (c) deploy in a non-intrusive way the assembly guideline; (d) assessing user’s progress in real-time by comparing user’s assembly with a model stored from the expert (including recognizing the parts, tools, and the current state of the assembly, which is a complex problem that fits in the object recognition area); (e) providing non-intrusive hint or feedback about the current and next state of the assembly, which is also a complex problem that aside the fact it also requires object recognition, it is needed to consider that there are many possible paths to achieve a correct assembly.
The purpose of this project was to develop an innovative Augmented Reality Product Assembly (ARPA) system able to:
- Implement the learn-by-example philosophy. The system developed learns while watching an expert human working on an assembly and will create the assembly guidelines capturing the assembly process followed by the expert human.
- Deploy virtual tutoring for novices. The system developed recognizes the process that a novice human is following trying to assemble pieces and will provide proper feedback.
Video – Last version of the system