An Application of a Digital Twin to Robotic System Design for an Unstructured Environment

摘要

Abstract An experimentable digital twin is created to aid in a design decision (beginning of life stage) for a robotic system. This product is meant to automate a material-feed system. The robot comprises a six-axis manipulator mounted on a mobile base. Due to variability in the dimensions of the material-feed system and positioning error of the mobile base, the material-placement routine is considered to take place in an unstructured environment. Working therein requires exteroceptive sensors, in this instance taking the form of computer vision. Data from this subsystem are used to match the geometry of the digital twin to the physical environment. This close correspondence between physical and virtual embodiments allows for significant design decisions to be reached from simulated experiments. In this case, two motion-planning approaches are compared and it is determined that the costs associated with implementing the dynamic one in the lab for testing are merited by its ease of use and reliability, since simulation-based control employs all current information.