Receding Hamiltonian-Informed Optimal Neural Control and State Estimation for Closed-Loop Dynamical Systems

TL;DR

This paper introduces Hamiltonian-Informed Optimal Neural controllers that leverage neural networks and Pontryagin's Maximum Principle for improved control and state estimation in dynamical systems, offering superior performance over existing methods.

Contribution

It presents a novel class of neural controllers, Hion, and the T-mano architecture, enabling customizable transient behavior, predictive control, and enhanced closed-loop feedback.

Findings

Hion controllers outperform traditional model-predictive controllers in optimality and tracking.

The framework effectively handles both linear and non-linear systems.

Comparative analysis shows superior control performance with the proposed methods.

Abstract

This paper formalizes Hamiltonian-Informed Optimal Neural (Hion) controllers, a novel class of neural network-based controllers for dynamical systems and explicit non-linear model-predictive control. Hion controllers estimate future states and develop an optimal control strategy using Pontryagin's Maximum Principle. The proposed framework, along with our Taylored Multi-Faceted Approach for Neural ODE and Optimal Control (T-mano) architecture, allows for custom transient behavior, predictive control, and closed-loop feedback, addressing limitations of existing methods. Comparative analyses with established model-predictive controllers revealed Hion controllers' superior optimality and tracking capabilities. Optimal control strategies are also demonstrated for both linear and non-linear dynamical systems.