Real-Time Deep-Learning-Driven Parallel MPC

A novel real-time approximated MPC control policy based on deep learning is proposed to address the high computational burden of model predictive control (MPC) for large-scale systems and those with fast dynamics. This control method approximates the optimal solution of the distributed optimization problems in the ALADIN-based parallel MPC design framework, resulting in a highly effective approach that outperforms other well-known methods for solving the MPC design problem. The numerical case study shows promising results, demonstrating the potential of this approach for real-time implementation.