English

Learning Accurate Long-term Dynamics for Model-based Reinforcement Learning

Machine Learning 2021-09-02 v2 Robotics

Abstract

Accurately predicting the dynamics of robotic systems is crucial for model-based control and reinforcement learning. The most common way to estimate dynamics is by fitting a one-step ahead prediction model and using it to recursively propagate the predicted state distribution over long horizons. Unfortunately, this approach is known to compound even small prediction errors, making long-term predictions inaccurate. In this paper, we propose a new parametrization to supervised learning on state-action data to stably predict at longer horizons -- that we call a trajectory-based model. This trajectory-based model takes an initial state, a future time index, and control parameters as inputs, and directly predicts the state at the future time index. Experimental results in simulated and real-world robotic tasks show that trajectory-based models yield significantly more accurate long term predictions, improved sample efficiency, and the ability to predict task reward. With these improved prediction properties, we conclude with a demonstration of methods for using the trajectory-based model for control.

Keywords

Cite

@article{arxiv.2012.09156,
  title  = {Learning Accurate Long-term Dynamics for Model-based Reinforcement Learning},
  author = {Nathan O. Lambert and Albert Wilcox and Howard Zhang and Kristofer S. J. Pister and Roberto Calandra},
  journal= {arXiv preprint arXiv:2012.09156},
  year   = {2021}
}

Comments

8 pages, +4 pages appendix

R2 v1 2026-06-23T21:01:39.253Z