@subheading(Figure 1a:)

	This and the following diagram are plan views of the cart's
world model at obstacle run stopping positions.

	In this first view, one step into a run, the cart has (by the
program's reckoning) travelled a little under a meter. The program has
noted a number of objects; a chair in the foreground, pieces of
cardboard on the floor, an icosahedron, a refrigerator, a cardboard
tree, and some other things.  It has deduced that the chair, the
icosahedron and the tree are real obstacles, and has planned a path
which slaloms around them (the cardboard in the foreground has been
ignored because it lies on the floor).

	The grid cells are two meter squares, conceptually on the
floor.  The cart's own position is indicated by the small heavy
square, and by the graph, indicating height, calibrated in
centimeters, to the left of grid. Since the cart never actually leaves
or penetrates the floor, this graph provides an indication of the
overall accuracy.  The irregular, tick marked, line behind the cart's
position is the past itinerary of the cart as deduced by the program.
Each tick mark represents a stopping place.  The picture at top of the
diagrams is the view seen by the TV camera. The two rays projecting
forward from the cart position show the horizontal boundaries of the
camera's field of view (as deduced by the camera calibration program).
The numbered circles in the plan view are features located and tracked
by the program.  The centers of the circles are the vertical
projections of the feature positions onto the ground. The size of each
circle is the uncertainty (caused by finite camera resolution) in the
feature's position. The length of the 45 degree line projecting to the
upper right, and terminated by an identifying number, is the height of
the feature above the ground, to the same scale as the floor grid.
The features are also marked in the camera view, in the guise of
numbered boxes. The thin line projecting from each box to a lower blob
is a stalk which just reaches the ground, in the spirit of the 45
degree lines in the plan view.  The irregular line radiating forwards
from the cart is the planned future path. This changes from stop to
stop, as the cart fails to obey instructions properly, and as new
obstacles are detected. The small ellipse a short distance ahead of
the cart along the planned path is the planned position of the next
stop.

@subheading(Figure 1b:)

	After the 11'th lurch the cart has rounded the chair, the
icosahedron and is planning to round the cardboard tree.  The wold
model has suffered som accumulated drift error, and the oldest
acquired features are considerably misplaced.  Also misplaced, due to
human error, is the cart's destination, which is about a meter behind
the far wall of the room.  The cart backed up a number of times later
in this run trying to get around it, when it found planned paths
blocked by newly observed portions of the wall.

The room

and the cart

The calibration pattern

Outdoors

Calibration-polynomial distorted grid superimposed on the spot
image from which it was calculated.