Saved in:
Bibliographic Details
Main Authors: Das, Shreya, Kumar, Kundan, Iqbal, Muhammad, Savolainen, Outi, Baumann, Dominik, Ruotsalainen, Laura, Särkkä, Simo
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2603.08468
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914380063440896
author Das, Shreya
Kumar, Kundan
Iqbal, Muhammad
Savolainen, Outi
Baumann, Dominik
Ruotsalainen, Laura
Särkkä, Simo
author_facet Das, Shreya
Kumar, Kundan
Iqbal, Muhammad
Savolainen, Outi
Baumann, Dominik
Ruotsalainen, Laura
Särkkä, Simo
contents Model-based reinforcement learning (MBRL) is sample-efficient but depends on the accuracy of the learned dynamics, which are often modeled using black-box methods that do not adhere to physical laws. Those methods tend to produce inaccurate predictions when presented with data that differ from the original training set. In this work, we employ Lagrangian neural networks (LNNs), which enforce an underlying Lagrangian structure to train the model within a Dyna-based MBRL framework. Furthermore, we train the LNN using stochastic gradient-based and state-estimation-based optimizers to learn the network's weights. The state-estimation-based method converges faster than the stochastic gradient-based method during neural network training. Simulation results are provided to illustrate the effectiveness of the proposed LNN-based Dyna framework for MBRL.
format Preprint
id arxiv_https___arxiv_org_abs_2603_08468
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Integrating Lagrangian Neural Networks into the Dyna Framework for Reinforcement Learning
Das, Shreya
Kumar, Kundan
Iqbal, Muhammad
Savolainen, Outi
Baumann, Dominik
Ruotsalainen, Laura
Särkkä, Simo
Systems and Control
Machine Learning
Model-based reinforcement learning (MBRL) is sample-efficient but depends on the accuracy of the learned dynamics, which are often modeled using black-box methods that do not adhere to physical laws. Those methods tend to produce inaccurate predictions when presented with data that differ from the original training set. In this work, we employ Lagrangian neural networks (LNNs), which enforce an underlying Lagrangian structure to train the model within a Dyna-based MBRL framework. Furthermore, we train the LNN using stochastic gradient-based and state-estimation-based optimizers to learn the network's weights. The state-estimation-based method converges faster than the stochastic gradient-based method during neural network training. Simulation results are provided to illustrate the effectiveness of the proposed LNN-based Dyna framework for MBRL.
title Integrating Lagrangian Neural Networks into the Dyna Framework for Reinforcement Learning
topic Systems and Control
Machine Learning
url https://arxiv.org/abs/2603.08468