Figure 22.

This schematic shows the adaptation of the algorithm by Antonelo et al. [2021] to the Oschmann et al. model and in combination with the deep learning algorithm initially developed by Yazdani et al. [2020]. The input of the neural network is expanded with a control input (Glu) and initial conditions $({\overrightarrow{x}}_o)$. The initial conditions of each interval are predicted by the neural network itself and replace the previously used concept of ${ℒ}_{Aux}$. AD stands for automatic differentiation.