Bone Hunt |
 |
Members
Daniel Munoz,
Siva Srinivasan
Summary
Goal:
- Have the dog start in unknown location in a known world and try to find
targets (EG: a bone) in a smart way. With only knowledge of the structure
of the world, the dog should be able to plan and navigate through the
pen to find targets and then figure out how to bring the target back to
where it started.
Accomplishments:
- Utilization of landmarks and particle filter to always know world location
- Created method for navigating to and from different landmarks in world shape space
- Rough motion sequence to find and pick up a bone from a different part of the pen
Future Work:
- Robustness. Need to reach landmarks in right position consistently
- Explore while building the map dynamically (and finding/retrieving targets at the same time)
Demo
Video demonstrating the Aibo's bone finding behavior:
Click to download video demo
Aibo beginning in one of the three starting locations
After localizing with starting pen, moving to the closest viewpoint
Going to the next viewpoint where the Aibo will search for a bone
The Aibo's location after picking up the bone
Returning to the viewpoint after grabbing the bone
Navigating between the two viewpoint positions on the way home
Coming back home to drop the bone off
