LAUSR

Currently, the automated assembly depends on advance programming, which is suitable to large-batch products assembly; however, it does not fit in the assembly of small-batch products due to the large amount of preparatory work including assembly planning and robot programming. Therefore, those assemblies in small batch largely rely on human interventions, which is a system-level problem. Targeting the problem, this paper presents a novel programming-free automated assembly planning and control approach based on virtual training. Within the context, the 3-D models of products are used, including general assembly features of each component. The features are used in a search-based planner to generate assembly sequence, and to plan assembly path. Then the virtual assembly simulation is carried out based on the generated assembly plan, where the collisions and contacts are captured and passed to the planner to regenerate a new path. The new path is simulated in the virtual world. The simulation process is repeated until an executable strategy is obtained. In the real world, the physical robots perform the actual assembly by following the trained sequences and paths that are calibrated according to the real positions and orientations of the components. A proof-of-concept case study is carried out in robot operating system environment to validate this approach.