A neural method for dynamical hierarchical optimal control of large-scale systems with multiple time delays

In the paper, a neural-network-based dynamic solver (LHCNN) is proposed for dynamical hierarchical optimal control of large-scale systems with multiple time delays. The local optimization subnetworks (LONN) and the coordination subnetwork (CNN) constituting the LHCNN can work simultaneously to provide the optimal solutions to the original problems. The LHCNN is globally stable and is inherently easy for implementation.