A framework for simulation-based optimization demonstrated on reconfigurable robot workcells

Today's trends towards automation and robotics, fueled by the emerging Industry 4.0 paradigm shift, open up many new kinds of control and optimization problems. At the same time, advances in 3D simulation technology lead to ever-improving simulation models and algorithms in various domains, such as multi-body dynamics, kinematics, or sensor simulation. This development can be harnessed for Simulation- based Optimization (SBO), where optimization results can be directly transferred from simulation models to the real world. In this paper, we introduce a formalism and modular framework for model configuration and SBO. We demonstrate the capabilities of our framework by optimizing the sensor layout within a reconfigurable robot workcell from the H2020 project ReconCell, allowing engineers to experiment with different optimizers and parameters. Evaluation of the results proves the usefulness of our approach and shows that the framework can be applied to a wide range of optimization problems without constraining the choice of simulation environment.