Expressive robot motions Robot Learning from Demonstration

This is a tutorial for introducing robot motions in ExpressiveRLfD project.

Welcome

Welcome to participate in this experiment. The whole experiment consists of 1+3 sessions and each takes ~3min to finish. The first session an introductory session for you to get familar with the environment. In the following 3 sessions, you need to teach the robot arm how to place the cup the designated area.


Task domain

Teaching robot to move the cup ( from ANY positions on the table ) to the green area.

Requirement for robot

  • No collisions between the cup and the boxes
  • Limited controls of cup (via keyboard arrow keys)
  • Must be within the table area

Requirement for you

  • 9 demonstration trials
  • 3 different cup positions
  • ~ 20 mins needed
Responsive image
Cup, table and boxes
Responsive image
Robot arm
Responsive image
Simulation environment

Robot motions

Three types of motions will be observed in the experiments:

Follow-up motion

The follow-up motion is to show that the robot is following up and learning from your actions.

Shaking motion

The shaking motion is to show the status when the robot is unsure about taking actions. Based on your demonstration, the robot will predict the next action and pause on the area where it is unsure about take more actions.

Nodding motion

The nodding motion is to show the status when the robot is very sure about taking actions based on its learning process. Nodding over the green area means the robot has found the target area while nodding over other areas indicates the robot mistakenly targets the designated area.


For any questions, please contact Alex Sun.