Solved Maze Memory

[acbuynak]

Once a maze is solved a few times, it is unnecessary to continually re-solve the maze just to update the position of the maze relative to the camera. A potential solution could be to separate vision capabilities:

• maze board tracking
• maze image segmentation, maze solver, and path registration
• tool path planner

Once segmented, all three components would be utilized to find and solve the maze initially. After a few successful solves, it is safe to assume the maze is known and will not change (unless a new maze is introduced). The successful solves could average the resulting local path and be fed into a tool path planner.

Assuming the maze doesn't change, the board tracking and tool path planner could be called to update the path as the maze moves around in the environment, but computation time not wasted on image segmentation and solving.
If deployed, the image segmentation should only be called periodically to ensure the maze hasn't changed.