Jove
Visualize
Contact Us

Related Experiment Videos

Autonomous learning based on cost assumptions: theoretical studies and experiments in robot control.

C H Ribeiro1, E M Hemerly

  • 1Technological Institute of Aeronautics, São José Dos Campos - SP, Brazil.

International Journal of Neural Systems
|May 8, 2000
PubMed
Summary

This study introduces a generalized Q-learning method for autonomous learning systems. It enables efficient experience acquisition by spreading information, improving reinforcement learning in robot navigation.

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

Adaptive observers for unknown general nonlinear systems.

IEEE transactions on systems, man, and cybernetics. Part B, Cybernetics : a publication of the IEEE Systems, Man, and Cybernetics Society·2008
Same author

Adaptive control for mobile robot using wavelet networks.

IEEE transactions on systems, man, and cybernetics. Part B, Cybernetics : a publication of the IEEE Systems, Man, and Cybernetics Society·2008
Same author

[Meningiomas of the cavernous sinus: the surgical resectability and complications].

Arquivos de neuro-psiquiatria·2001
Same author

[Subfrontal approach in sellar and suprasellar lesions].

Arquivos de neuro-psiquiatria·2000
Same author

Autonomous learning based on cost assumptions: theoretical studies and experiments in robot control.

International journal of neural systems·1999
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

Area of Science:

  • Artificial Intelligence
  • Machine Learning
  • Robotics

Background:

  • Autonomous learning relies on experience acquisition, which is often time-consuming in realistic applications.
  • Classical learning algorithms struggle with the crudeness of information from sensor readings, actuator control, and algorithmic updates.
  • There's a need to embed facilitating information into learning processes without compromising fundamental algorithm structures.

Purpose of the Study:

  • To investigate a generalized formulation of the Q-learning method.
  • To enable the spreading of information from single updates to neighboring states, facilitating faster convergence.
  • To demonstrate a mechanism for safely embedding prior knowledge into the state space structure.

Main Methods:

  • Developed a generalized Q-learning formulation that allows information propagation to neighboring states.

Related Experiment Videos

  • Modified a reinforcement learning algorithm to incorporate this generalized Q-learning approach.
  • Tested the modified algorithm in a real-world robot navigation task.

    • Main Results:

    • The generalized Q-learning method converges to optimality by spreading information.
    • The formulation effectively embeds prior knowledge about the state space structure.
    • Demonstrated successful application in a practical robot navigation scenario.

    Conclusions:

    • The proposed generalized Q-learning method enhances autonomous learning efficiency.
    • This approach offers a robust way to integrate prior knowledge into reinforcement learning.
    • The findings are validated through a real robot navigation task, showing practical applicability.