Robust Real-Time Model Predictive Control for Industrial Manipulators via Learned Dynamics

DOI

10.2991/978-94-6463-980-3_30How to use a DOI?

Keywords

Model Predictive Control; Data-Driven Control; Robot Manipulation; Real-Time Systems; Machine Learning for Robotics

Abstract

Real-time Model Predictive Control (MPC) based on analytical models has emerged as a powerful technique for the high-frequency control of robotic manipulators. However, the performance of these controllers is fundamentally limited by the fidelity of their underlying models, often leading to degraded performance when faced with unmodeled dynamics such as varying payloads. In this paper, we address this limitation by proposing a novel Data-Driven Constrained Quadratic MPC (DDCQ-MPC) framework. Our key idea is to replace the traditional analytical model with a neural network that learns the complex, real- world dynamics of the robot directly from trajectory data. This learned model is then linearized in real-time using automatic differentiation, allowing us to retain the fast, convex optimization structure of a Quadratic Program (QP) that can be solved at 1 kHz. We validate our approach through a series of experiments on large-scale, real-world datasets. The results demonstrate that our data-driven controller significantly outperforms a state-of-the-art analytical MPC baseline in tasks involving heavy, unmodeled payloads, exhibiting substantially lower tracking errors and higher task success rates.

Copyright

© 2025 The Author(s)

Open Access

Open Access This chapter is licensed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International License (http://creativecommons.org/licenses/by-nc/4.0/), which permits any noncommercial use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.

Download article (PDF)

TY  - CONF
AU  - Jiwen Yang
PY  - 2025
DA  - 2025/12/26
TI  - Robust Real-Time Model Predictive Control for Industrial Manipulators via Learned Dynamics
BT  - Proceedings of the 2025 5th International Conference on Business Administration and Data Science (BADS 2025)
PB  - Atlantis Press
SP  - 321
EP  - 329
SN  - 2352-538X
UR  - https://doi.org/10.2991/978-94-6463-980-3_30
DO  - 10.2991/978-94-6463-980-3_30
ID  - Yang2025
ER  -