Learning from Demonstration for Use with Teleoperation

1
Learning from Demonstration for Use with
Teleoperation

• Josh Southerland
• Charles de Granville

2
Our Workspace
3
Error in position and orientation to three goal
objects.
Finger flexion information from the P5 Data Glove.
4
Error in position and orientation to three goal
objects.
Velocity data. Meters/second plotted against
time steps.
5
Velocity(blue) scaled for comparison with the
error data for each goal
Red grab
Black grab
Green grab
6
Objectives

• Create a system capable of parsing out which
  operations actually took place.
• Develop that system to handle changes to the
  environment.
• Apply the system as a form of Supervised Learning
  for robots.