Computationally efficient Gauss-Newton reinforcement learning for model predictive control

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Brandner, Dean, Gros, Sebastien, Lucia, Sergio
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910095460270080
author Brandner, Dean
Gros, Sebastien
Lucia, Sergio
author_facet Brandner, Dean
Gros, Sebastien
Lucia, Sergio
contents Model predictive control (MPC) is widely used in process control due to its interpretability and ability to handle constraints. As a parametric policy in reinforcement learning (RL), MPC offers strong initial performance and low data requirements compared to black-box policies like neural networks. However, most RL methods rely on first-order updates, which scale well to large parameter spaces but converge at most linearly, making them inefficient when each policy update requires solving an optimal control problem, as is the case with MPC. While MPC policies are typically low parameterized and thus amenable to second-order approaches, existing second-order methods demand second-order policy derivatives, which can be computationally intractable. This work introduces a Gauss-Newton approximation of the deterministic policy Hessian that eliminates the need for second-order policy derivatives, enabling superlinear convergence with minimal computational overhead. To further improve robustness, we propose a momentum-based Hessian averaging scheme for stable training under noisy estimates coupled with an adaptive trustregion. We demonstrate the effectiveness of the approach on a nonlinear continuously stirred tank reactor (CSTR), showing faster convergence and improved data efficiency over state-of-the-art firstorder methods and deep RL approaches.
format Preprint
id arxiv_https___arxiv_org_abs_2508_02441
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Computationally efficient Gauss-Newton reinforcement learning for model predictive control
Brandner, Dean
Gros, Sebastien
Lucia, Sergio
Systems and Control
Machine Learning
Model predictive control (MPC) is widely used in process control due to its interpretability and ability to handle constraints. As a parametric policy in reinforcement learning (RL), MPC offers strong initial performance and low data requirements compared to black-box policies like neural networks. However, most RL methods rely on first-order updates, which scale well to large parameter spaces but converge at most linearly, making them inefficient when each policy update requires solving an optimal control problem, as is the case with MPC. While MPC policies are typically low parameterized and thus amenable to second-order approaches, existing second-order methods demand second-order policy derivatives, which can be computationally intractable. This work introduces a Gauss-Newton approximation of the deterministic policy Hessian that eliminates the need for second-order policy derivatives, enabling superlinear convergence with minimal computational overhead. To further improve robustness, we propose a momentum-based Hessian averaging scheme for stable training under noisy estimates coupled with an adaptive trustregion. We demonstrate the effectiveness of the approach on a nonlinear continuously stirred tank reactor (CSTR), showing faster convergence and improved data efficiency over state-of-the-art firstorder methods and deep RL approaches.
title Computationally efficient Gauss-Newton reinforcement learning for model predictive control
topic Systems and Control
Machine Learning
url https://arxiv.org/abs/2508.02441