Optical data constitutes a valuable tool for navigation.
Tanner Labs has developed Cartesian Flow, an efficient
algorithm for computing optical flow, and demonstrated, using
theoretical analysis and empirical simulations, that it can
be used for robotic self-navigation in a variety of environments.
Then, we implemented Cartesian Flow by developing a reconfigurable
computer that offers higher performance than a custom IC.
We based the reconfigurable computer design on our neural
network reconfigurable computer; they share the same motherboard
and communications infrastructure. Its flexible and expandable
format can be easily modified to include more sophisticated
optical flow computations or post-flow navigational algorithms
(three more daughtercards can be added).
This research has been sponsored by the U.S. Department of
Commerce under the SBIR