Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Experiment Videos

Robust reinforcement learning.

Jun Morimoto1, Kenji Doya

  • 1Computational Brain Project, ICORP, JST, Sora Ku-gun, Kyoto 619-0288, Japan. xmorimo@atr.jp

Neural Computation
|February 22, 2005
PubMed
Summary
This summary is machine-generated.

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by
Same author

Towards natural stand-up movement support: guiding higher-dimensional muscle activation using a Lower-DOF assistive chair.

Frontiers in bioengineering and biotechnology·2026
Same author

Rapid functional reorganization of the targeted contralesional hemisphere induced by one week of noninvasive closed-loop neurofeedback guides motor recovery in post-stroke patients with chronic motor impairment: a phase I trial.

Communications medicine·2026
Same author

A computational model of canonical cortical microcircuits for dynamic Bayesian inference and control as inference.

Neuroscience research·2025
Same author

Dynamical modeling of torso stability in running via hip-knee three pairs of six springs.

Bioinspiration & biomimetics·2025
Same author

Neural-enhanced motion-to-EMG: refining simulated muscle activity from musculoskeletal models using a Seq2Seq approach.

Frontiers in bioengineering and biotechnology·2025
Same author

Possible contribution to data-driven primate research: Comment on "Kinematic coding: Measuring information in naturalistic behaviour" by Becchio, Pullar, Scaliti, and Panzeri.

Physics of life reviews·2025
Same journal

A Model-Free Reinforcement Learning Implementation of Decision Making Under Uncertainty by Sequential Sampling.

Neural computation·2026
Same journal

DROP: Distributional and Regular Optimism and Pessimism for Reinforcement Learning.

Neural computation·2026
Same journal

Hierarchical Active Inference Using Successor Representations.

Neural computation·2026
Same journal

W-Kernel and Its Principal Space for Frequentist Evaluation of Bayesian Estimators.

Neural computation·2026
Same journal

A Hidden Markov Model-Inspired Sequence Classification Method for Hyperdimensional Computing.

Neural computation·2026
Same journal

Sparse Graphical Modeling for Electrophysiological Phase-Based Connectivity Using Circular Statistics.

Neural computation·2026
See all related articles

This study introduces robust reinforcement learning (RRL), a new method addressing model errors in reinforcement learning (RL). RRL enhances control system stability and performance by accounting for environmental disturbances and inaccuracies.

Area of Science:

  • Robotics
  • Control Theory
  • Artificial Intelligence

Background:

  • Reinforcement learning (RL) commonly uses environmental models for offline and online tasks.
  • Discrepancies between models and real environments can cause unpredictable outcomes.
  • Existing RL methods struggle with input disturbances and modeling errors.

Purpose of the Study:

  • Propose a novel reinforcement learning paradigm, robust reinforcement learning (RRL).
  • Explicitly incorporate input disturbance and modeling errors into RL.
  • Develop online learning algorithms for robust control.

Main Methods:

  • Formulate the problem as a differential game based on H(infinity) control theory.
  • Seek a min-max solution for a value function considering rewards and disturbance norms.

Related Experiment Videos

  • Derive online algorithms for value function estimation and optimal control/disturbance calculation.

    • Main Results:

    • RRL algorithms successfully learned policies and value functions matching analytical solutions in linear inverted pendulum tasks.
    • RRL demonstrated robust performance against changes in pendulum weight and friction in nonlinear tasks.
    • Standard RL algorithms failed to adapt to these changes, unlike RRL.

    Conclusions:

    • RRL provides a robust approach to reinforcement learning by mitigating the impact of model errors and disturbances.
    • The proposed method significantly improves control system performance in dynamic and uncertain environments.
    • RRL is effective for complex control tasks like the cart-pole swing-up.