Beyond Constant Parameters: Hyper Prediction Models and HyperMPC

Jan Węgrzynowski, Piotr Kicki, Grzegorz Czechmanowski, Maciej Piotr Krupka, Krzysztof Walas
Proceedings of The 9th Conference on Robot Learning, PMLR 305:4885-4907, 2025.

Abstract

Model Predictive Control (MPC) is among the most widely adopted and reliable methods for robot control, relying critically on an accurate dynamics model. However, existing dynamics models used in the gradient-based MPC are limited by computational complexity and state representation. To address this limitation, we propose the Hyper Prediction Model (HyperPM) - a novel approach in which we project the unmodeled dynamics onto a time-dependent dynamics model. This time-dependency is captured through time-varying model parameters, whose evolution over the MPC prediction horizon is learned using a neural network. Such formulation preserves the computational efficiency and robustness of the base model while equipping it with the capacity to anticipate previously unmodeled phenomena. We evaluated the proposed approach on several challenging systems, including real-world F1TENTH autonomous racing, and demonstrated that it significantly reduces long-horizon prediction errors. Moreover, when integrated within the MPC framework (HyperMPC), our method consistently outperforms existing state-of-the-art techniques.

Cite this Paper

BibTeX
@InProceedings{pmlr-v305-wegrzynowski25a, title = {Beyond Constant Parameters: Hyper Prediction Models and HyperMPC}, author = {W\k{e}grzynowski, Jan and Kicki, Piotr and Czechmanowski, Grzegorz and Krupka, Maciej Piotr and Walas, Krzysztof}, booktitle = {Proceedings of The 9th Conference on Robot Learning}, pages = {4885--4907}, year = {2025}, editor = {Lim, Joseph and Song, Shuran and Park, Hae-Won}, volume = {305}, series = {Proceedings of Machine Learning Research}, month = {27--30 Sep}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v305/main/assets/wegrzynowski25a/wegrzynowski25a.pdf}, url = {https://proceedings.mlr.press/v305/wegrzynowski25a.html}, abstract = {Model Predictive Control (MPC) is among the most widely adopted and reliable methods for robot control, relying critically on an accurate dynamics model. However, existing dynamics models used in the gradient-based MPC are limited by computational complexity and state representation. To address this limitation, we propose the Hyper Prediction Model (HyperPM) - a novel approach in which we project the unmodeled dynamics onto a time-dependent dynamics model. This time-dependency is captured through time-varying model parameters, whose evolution over the MPC prediction horizon is learned using a neural network. Such formulation preserves the computational efficiency and robustness of the base model while equipping it with the capacity to anticipate previously unmodeled phenomena. We evaluated the proposed approach on several challenging systems, including real-world F1TENTH autonomous racing, and demonstrated that it significantly reduces long-horizon prediction errors. Moreover, when integrated within the MPC framework (HyperMPC), our method consistently outperforms existing state-of-the-art techniques.} }
Endnote
%0 Conference Paper %T Beyond Constant Parameters: Hyper Prediction Models and HyperMPC %A Jan Węgrzynowski %A Piotr Kicki %A Grzegorz Czechmanowski %A Maciej Piotr Krupka %A Krzysztof Walas %B Proceedings of The 9th Conference on Robot Learning %C Proceedings of Machine Learning Research %D 2025 %E Joseph Lim %E Shuran Song %E Hae-Won Park %F pmlr-v305-wegrzynowski25a %I PMLR %P 4885--4907 %U https://proceedings.mlr.press/v305/wegrzynowski25a.html %V 305 %X Model Predictive Control (MPC) is among the most widely adopted and reliable methods for robot control, relying critically on an accurate dynamics model. However, existing dynamics models used in the gradient-based MPC are limited by computational complexity and state representation. To address this limitation, we propose the Hyper Prediction Model (HyperPM) - a novel approach in which we project the unmodeled dynamics onto a time-dependent dynamics model. This time-dependency is captured through time-varying model parameters, whose evolution over the MPC prediction horizon is learned using a neural network. Such formulation preserves the computational efficiency and robustness of the base model while equipping it with the capacity to anticipate previously unmodeled phenomena. We evaluated the proposed approach on several challenging systems, including real-world F1TENTH autonomous racing, and demonstrated that it significantly reduces long-horizon prediction errors. Moreover, when integrated within the MPC framework (HyperMPC), our method consistently outperforms existing state-of-the-art techniques.
APA
Węgrzynowski, J., Kicki, P., Czechmanowski, G., Krupka, M.P. & Walas, K.. (2025). Beyond Constant Parameters: Hyper Prediction Models and HyperMPC. Proceedings of The 9th Conference on Robot Learning, in Proceedings of Machine Learning Research 305:4885-4907 Available from https://proceedings.mlr.press/v305/wegrzynowski25a.html.

Related Material