06.1 - Flight Dynamics and Control (Control & Modelling)
INTEGRATING POST-OPTIMAL SENSITIVITIES INTO SUPERVISED TRAINING OF NEURAL NETWORKS
J. Diepolder¹, J.Z. Ben-Asher¹; ¹Technion, Israel
In this paper, we present an advancement for learning quantities derived from solutions to optimal control problems. The proposed approach leverages sensitivity information obtained at optimal solutions to enhance the supervised training of neural networks.