Learning the Binding Affordances of Objects: A Behavior-Grounded Approach

This paper introduces a developmental approach to learning the binding affordances of objects by a robot. A behavior-based framework is used to ground the affordance representation in the behavioral repertoire of the robot. The affordances are learned during a behavioral babbling stage in which the robot randomly chooses sequences of exploratory behaviors, applies them to the objects, and detects invariants in the resulting set of observations. The invariants are calculated relative to the robot’s body. The approach was implemented and tested in a dynamics robot simulator.

Download

Stoytchev, A., "Toward Learning the Binding Affordances of Objects: A Behavior-Grounded Approach", In proceedings of AAAI Symposium on Developmental Robotics, Stanford University, Mar 21-23, 2005.

BibTeX Entry

@InProceedings{Stoytchev2005,
  author =     {Alexander Stoytchev},
  title =      {Toward Learning the Binding Affordances of Objects: A Behavior-Grounded Approach},
  booktitle =  {Proceedings of AAAI Symposium on Developmental Robotics},
  year =       {2005},
  pages =      {17-22},
  address =    {Stanford University},
  month =      {Mar 21-23},
}

MPEG Movies from the Simulator

The videos show the same sequence of exploratory behaviors executed on three different objects.

• Spindle Object
• Mallet Object
• Dumbbell Object
• H-frame Object

Binding with an H-frame object using the open-gripper behavior.

• H-frame Object