Subsumption Agent: Seymour
Subsumption Agent: Seymour
The Seymour robot is designed with enough onboard computers
to support the vision processing of 8-10 low resolution
cameras at approximately 10 frames per second. The cameras feed different
subsumption layers so that one layer could attend to ceiling lights
while the other could look for obstacles in the path. So the final
robot system could have at least five different layers.
This system is still in the construction phase, but preliminary
testing of the vision algorithms were being performed on Allen with good results. Although these
routines are simple, noisy, and unreliable, the robot can track,
follow, and chase objects by switching between different simple
visual routines as
one fails. It is important to note that nowhere in the
subsumption program is there the
notion of an identifiable object
internally. But to an outside observer, it certainly appears to
follow a moving object very well.
Return to the top of this architecture.
Current Location: Subsumption-Capabilities-RoboticTasks-Seymour