Process Simulation Using Physical Joining Models For Learning Robot-Based Assembly Tasks

The ambitious requirements for the economical production of personalized products are one of the key drivers in the development of flexible automation solutions. Novel solution approaches for programming robots using advanced Machine Learning methods show promising results to increase flexibility. These so called intelligent automation solutions can adapt to changes in the product or manufacturing process with only minor human investment. However, the significant amounts of data needed to train the robots still limits the approach. One possible way is to generate this data is utilizing a physics simulation environment. Specialized joining models can be used to further increase the accuracy of the simulation and enrich the generated data as well as reduce the gap between the simulation and the real world. The presented work focussed on the extension of state of the art simulation environments with dedicated joining models for the assembly of electrical components in the electrical cabinet assembly. This extension is intended to make a significant contribution to Simulation-based training for intelligent robot-based automation solutions.