The Maze was constructed out of two 4 by 8-foot sheets of plywood bordered by 2 by 4-inch pine boards. The maze obstacles were fabricated into 6 by 6 by 6-inch cubes from standard rail ties.
The surface of the maze was primed and then painted a dark gray with polyurethane paint. Dark gray was chosen to simulate the relative absence of light value and thus simplify the vision module's task. The maze surface was then buffed with an abrasive floor buffer, which removed surface shine and significantly reduced any glare off the maze surface. Additionally, the abrasive buffing provided a superior surface for the traction of the robot's wheels.
Originally, the maze obstacles were painted with the same dark gray polyurethane paint and the surface of the cube was left rough and unfinished. Later, we found that the robot's infrared sensors would not consistently report the proximity of the rough wood surface. Consequently, gray paper was stapled to the sides of each obstacle, which provided us with a better obstacle surface for proximity sensing.
Originally, one-quarter inch wide, white typographer's tape was employed to segment the individual 256 cells on the maze surface. Later, these same lines were employed to prescribe a straight path, along which, the robot could travel. Employing these lines as paths effectively reduced the number of cells in the maze from 256 to 225. This change modified our goal of having a 256 cell maze. Additionally, this change required a small adjustment in the vision program so that it correctly reports the cell the robot occupies. At a future date and without rebuilding the entire maze, these lines may easily be removed and replaced to provide a maze of 256 cells which, also prescribe a path along which, the robot may travel.
The wooden outside border of the maze was painted a light gray in contrast to the dark maze surface and the white maze lines. Handles were attached to these borders to facilitate transportation of the maze.
Photo courtesy of Pat Leang
Used with permission.
Introduction | Overview | Maze | Interface | Learning | Vision | Robot | Integration | Bibliography