In a way the mathematical models inside the robot make it possible that the robot simulates the real world. This simulation is done with a network of equations, rather complex sets of equations with a lot of parameters. One could call these equations Functional-Mock-Ups, and the simulated Robot could be named a Digital Twin under certain circumstance, i.e. when the Digital Twin and the Robot are synchronized, making sure that simulation and reality do not differ too much (with a distance measure to be set by a user of both, the digital twin and the robot system). If this simulation is then used to check different options for the best next action of the robot, then the model knowledge is incorporated into the robot. If the simulation of the robot automatically choosing the best next action is used to optimize the robot design, the architecture can also be used to setup such extended simulations. (However, optimizing optimized robots with a modelled synchronization mechanism is a heavy workload for the simulating computer. But, at least, the simulation setup can be described.) Instead of performing simulations, we are able to solve the optimization equations analytically, as we are able to generate the Cascaded Gaussian Functionals as explained in the previous chapter. Hence, we are able to provide an automatic software generation from the simulation setup.