🚀 go-pugleaf

On Thursday, July 11, 2013 3:00:24 AM UTC+10, han...@bbs.cpcn.com wrote:
> The Jan 1969 issue of C&A had a "15 years ago" article that described research at Bell Labs using a computerized mouse to track its way through a maze.
> 
> 
> 
> The latest of Shannon's machines is a mouse which solves a
> 
> maze, as countless mice do in psychology laboratories throughout
> 
> the world. The difference is that Shannon's mouse is a
> 
> realistic, lifesize lump of gray rubber, molded around a magnet,
> 
> adorned with copper whiskers, and mobilized on three
> 
> wheels. As a maze-solver it is definitely in the super-mouse
> 
> class.
> 
> 
> 
> The maze is about half the size of a desk top and has
> 
> aluminum fences which can be slipped into 40 different slots
> 
> in order to create problems for the mouse. When Shannon
> 
> places the mouse at any spot in the maze, it starts scurrying
> 
> for a piece of cheese (a brass contact which rings a bell) in
> 
> one corner. Every time the animal runs head on into a fence,
> 
> it backs away, makes a quarter turn, and scuttles off in a new
> 
> direction. By this trial and error process it finally stumbles
> 
> upon the cheese. Then comes the truly extraordinary performance.
> 
> 
> 
> Shannon replaces the mouse in some part of the
> 
> maze. This time it proceeds surely and swiftly toward the
> 
> cheese, never so much as brushing against a fence. It has
> 
> learned the maze. If Shannon then alters the maze, the mouse
> 
> resorts to further trial and error until it has mastered the fresh
> 
> arrangement of fences.
> 
> 
> 
> Naturally, this ability to learn a complicated pattern and to
> 
> relearn its lesson when the pattern changes cannot yet be
> 
> crammed into so tiny a machine as the mouse's body. The
> 
> brain, indeed, is in a separate case.
> 
> 
> 
> Stripped of its fetching animism, this is the way the mouse
> 
> works: When it is set on the metal floor of the maze, its
> 
> magnet trips a switch under the floor. A motor-driven electromagnet
> 
> hurries to a spot directly beneath the mouse and
> 
> seizes it in a magnetic grip. The magnet turns the mouse 90
> 
> degrees and propels it forward. If the copper whiskers make
> 
> contact with a fence, closing a circuit, the magnet backs
> 
> away, turns the mouse again, and tries anew to move it forward.
> 
> Each successful sally is recorded in the mouse's brain,
> 
> which is controlled by 50 relays. Interpretively, this brain
> 
> considers the maze as a number of square sections. When the
> 
> mouse moves without interference from one section to another,
> 
> the brain registers in its memory the equivalent of: "turn
> 
> North in section 5." By the time the mouse has visited every
> 
> section, its brain has learned how to steer it clear of all blind
> 
> alleys.
> 
> 
> 
> Important as they may be to telephone engineers, Shannon's
> machines may eventually prove to be even more significant in
> the life of the average citizen - the man who knows little
> about telephones beyond how to use them. For Shannon's
> machines promise to be progenitors of a new breed of supergadgets,
> using the control principles which he is working out.
> From the mouse it is only a short step of engineering imagination
> to an automatic lawnmower which could be instructed
> to pick its way through an ornamental garden, clipping the
> grass without ever blundering across a flower bed. Another
> might retrieve balls for tennis players. Still another might
> mix and bake a cake from a selection of factory-tested recipes.

A mouse in a maze was programmed on DEUCE in about 1958.
The maze, cheese, and mouse, were displayed on the computer's VDU.

In the first trip thru the maze, the mouse tried virtually every
path (including the dead-ends), as it sought to find the cheese.

For the second trip, the mouse -- having learned the route --
went straight to the cheese, without entering any of the dead-ends.