Autonomous Robot Skill Acquisition - AAAI 2011 Best Student Video

Loading...

Sign in or sign up now!
Alert icon
Upgrade to the latest Flash Player for improved playback performance. Upgrade now or more info.
815 views
Loading...
Alert icon
Sign in or sign up now!
Alert icon

Uploaded by on Aug 9, 2011

This video demonstrates autonomous skill acquisition on a robot acting in a constrained environment called the "Red Room." The environment consists of buttons, levers, and switches, all located at points of interest designated by ARTags. The robot can navigate to these areas and perform primitive manipulation actions, some of which affect the physical state of the maze (by opening or closing a door).

In the first experiment, the robot must learn to sequence the primitive controllers available to solve the task which is to push the green switch in the second room. To do this, the robot must 1) press the button, 2) pull the lever (opening the door), 3) navigate to the second room, and 4) push the switch. Once the policy is learned, the robot automatically segments the learned optimal trajectory into component skills with corresponding state and motor abstractions.

To test the effectiveness of the extracted skills, we built a new Red Room and evaluated the robot's performance in the early stages of learning. In the first case, the robot only had access to the original set of primitive controllers. In the second case, the robot also had access to the extracted manipulation skills.

For more details, please see:

G.D. Konidaris, S.R. Kuindersma, R.A. Grupen, A.G. Barto. Autonomous Skill Acquisition on a Mobile Manipulator. In Proceedings of the Twenty-Fifth Conference on Artificial Intelligence (AAAI-11), pages 1468-1473. San Francisco, CA, August 2011.

Category:

Science & Technology

Tags:

License:

Standard YouTube License

  • likes, 0 dislikes

Link to this comment:

Share to:

Uploader Comments (skuindersma)

  • So what programming language did u use for the robot?am starting on prolog and C but this is a target experiment i am looking forward to.Quite intresting the how intellingent that system is.

  • @myxxmichuki The robot's base API is written in C++, but we used Microsoft Robotics Studio (using C#) to integrate learning, perception, and control processes across multiple machines. The embedded processor for the Red Room environment was programmed in C.

    We have since converted the robot's code base to ROS, so all of our applications are written in C++ or Python.

see all

All Comments (3)

Sign In or Sign Up now to post a comment!
  • Do you know who were nominated for/won the other awards?

Loading...

Alert icon
0 / 00Unsaved Playlist Return to active list
    1. Your queue is empty. Add videos to your queue using this button:
      or sign in to load a different list.
    Loading...Loading...Saving...
    • Clear all videos from this list
    • Learn more