Assumptions
Through design and implementation, I have noticed that I rely
substantially on some environmental factors in order to operate
correctly. Also, due to my Player/Stage implementation, I make a
handful of other assumptions when translating my approach to the real
world.
- Easy Robot Differentiation: the fiduciary device allows for a
unique ID, but the real world makes no such guarantees.
- Sensor Accuracy: I assume that my laser readings have zero jitter.
If we took this assumption away, every time the observer moves, we
would suffer substantial additional error.
- Initial Rover
Proximity: My implementation requires robots to be placed next to each
other prior to execution. This is a reasonable assumption.
- Static World Assumption: Though the occupancy grid paradigm can
account for new unexpected data, the way I handle a robot leaving its
observer's view is simply to have the Scout back up. If a door were
to close between them unexpectedly, backing up would not suffice.
- No Kidnapping of Robots: If the static observer were to be moved
unexpectedly but still can see its Scout and does not detect failure,
the map would offset itself and we'd all be screwed. I've tried
moving him; it's not pretty.
Given the nature and medium of this project, I don't think these are
outlandish assumptions.
All
material & intellectual property on this site
was gathered by
Scott Jason Friedman for CS553
at Washington University in
St. Louis, 2004.
|