Draft:Moonlight Special Micromouse
Moonlight Special Micro-Mouse 1978-1979
From Wikipedia, the free encyclopedia
Moonlight Special was a pioneering self-contained autonomous mobile robot developed by engineers Art Boland, Ron Dilbeck, and Phil Stover at the Pacific Northwest National Laboratory (PNNL) in Richland, Washington, under the management of the Battelle Memorial Institute. It is recognized as a milestone in the development of self-contained robots capable of simultaneous localization and mapping, first demonstrated at the National Computer Conference in June 1978 (Anaheim, CA).[1][2][3]
 | Review waiting, please be patient.
This may take 2 months or more, since drafts are reviewed in no specific order. There are 3,847 pending submissions waiting for review.
Where to get help
How to improve a draft
You can also browse Wikipedia:Featured articles and Wikipedia:Good articles to find examples of Wikipedia's best writing on topics similar to your proposed article. Improving your odds of a speedy review To improve your odds of a faster review, tag your draft with relevant WikiProject tags using the button below. This will let reviewers know a new draft has been submitted in their area of interest. For instance, if you wrote about a female astronomer, you would want to add the Biography, Astronomy, and Women scientists tags. Editor resources
Reviewer tools
|
| Manufacturer | Pacific Northwest National Laboratory (Battelle) |
|---|
The robot was a primary Micromouse competitor in the "Amazing Micro-Mouse Maze Contest" (1978–1979), a seminal event in robotics history sponsored by IEEE Spectrum magazine. The competition, announced in 1977, challenged engineers to build a robot capable of escaping a maze without external control and included an entry form in the January 1978 issue of the magazine. However, as explained in a later issue of IEEE Spectrum there was a loophole, "The challenge was to employ microprocessor technology to deisgn and construct a self-contained "thinking mouse" that could solve a maze and, in subsequent trials, avoid its earlier mistakes. A loophole in the rules [outside wall exit], however, let strictly mechanical, "nonintelligent" [ wall follower ] mice enter, too."[2][4][5]
After Moonlight Special's success in the first Contest Time Trials in June, 1978, the team's previously personal off-hours (moonlighting) project gained Battelle Engineering Physics department status, a modest parts budget and access to additional internal resources. The Graphics department created an eye-catching 'classic mouse grey' fiberglass body, including whimsical details such as fuzzy ears with pink satin lining, beady black eyes, a pink nose, black whiskers and tail, and a 'wind-up' key on top that was used to easily lift the robot out of a maze. By October 1978, the project had significant internal recognition, featuring prominently during a visit by the Battelle Memorial Institute Board of Trustees. Moonlight Special was examined by BMI President Sherwood Fawcett and Trustee John R. Pierce, a notable pioneer in communications technology.[6]
Technical Design
Built from scratch, Moonlight Special utilized a Zilog Z80A microprocessor and operated within a constrained 1 KB RAM plus 4 KB eraseable PROM environment. Its architecture was notable for separating low-level clock-driven stepper motor control and sensor-reading interrupts from high-level cognitive navigation logic. The stepper motors enabled distance measurement, straight line travel and precise turning maneuvers; keys to avoiding wall contact. Two 45 degree angle look-back-across-the-wall-tops photodiode sensors provided bang-bang control for straight line travel and post-turn-maneuver wall alignment corrections. Three more look-back photodiode sensors (two right angle sides and one front) detected left/right wall openings and dead ends. Two battery packs provided on-board power. The microprocessor/photodiode pack used four rechargeable AA batteries (nominal 5V), and the motor/infrared LED light source battery pack used ten rechargeable Sub-C batteries (nominal 12V). Run time was 25 minutes per charge cycle. All on-board software was coded in Z80 assembly language and cross-assembled using a PDP-11/70.[1][2]
- Maze Mapping: The system utilized dead reckoning and Graph theory to map in memory up to 32 nodes within a contest maze. As documented in the original technical proceedings, each node table entry included linkage and distance storage for all four compass directions in a N/W/S/E orientation, relative to the starting position. This structured adjacency list allowed the robot to navigate 3-way or more complex 4-way intersections and arbitrary geometries.[1]
- Sensor Modulation: To overcome 'blinding' caused by exhibit and TV news camera lighting during the 1978 Time Trials (initially solved by adding a light blocking hood), the team implemented signal modulation for its infrared LED light sources and photodiode sensors, a critical refinement for robust autonomous navigation.[1]
Navigation Logic
While contemporary entries often relied on simple wall-hugging heuristics, Moonlight Special utilized a multi-pass learning strategy, validated prior to robot design with a FORTRAN simulation using a Tektronix 4010 vector graphics display terminal:[7]
.jpg)
- Discovery: The software used an adaptation of the Tarry Algorithm (Gaston Tarry 1895) to ensure exhaustive and efficient exploration of unknown mazes for exit discovery and maze-mapping. Exit discovery was accomplished by logging the most recently visited node's coordinates and departure direction when the robot's "Run/Stop" switch was moved to Stop during first maze runs, when the robot's exploration path reached the exit. This generalized approach worked for exterior exits as well as interior goal locations.[1][8]
- Virtualization: The software logic used virtual "wall blocking codes" in RAM to virtually mark dead ends (first and second runs) and to virtually close the exit during second runs, forcing the robot to complete its internal maze map by the end of second runs, and return to the entrance.[1][8]
- Optimization: For final runs in 1979, the robot utilized its internal maze map (Graph) and Virtual Depth-First Search (backtracking) to calculate the mathematically shortest path in memory, and then used a new set of "wall blocking codes" to force the robot to traverse the shortest path from maze entrance to exit. The final run could be repeated multiple times to demonstrate that the shortest path was well known in the robot's memory.[1][9][3]
Competition History
The Battelle project demonstrated high software stability, with the core navigation logic shared across multiple hardware platforms during the Amazing Micro-Mouse Maze Contest.[2]
- 1978 National Computer Conference (NCC) Time Trials (Anaheim): Moonlight Special debuted, demonstrating its ability to completely map a maze and utilize "dead-end avoidance" to improve run times. The runs were described in OMNI magazine, "Moonlight Special began its test run by exploring every blind alley of the one-and-a-half-by-two-and-a-half-meter maze. Then, on the third try, Moonlight computed the best route and rolled directly to the exit - in a sizzling 51.4 seconds. To top it off, on a "victory run" after the competition, Moonlight flashed through the labyrinth again in 40 seconds flat to demonstrate how much it had learned from the previous trip." [1][2][3]
Moonlight Special (with mouse body removed) at the IEEE Spectrum Amazing Micro-Mouse Maze Contest finals, New York, NY, June 1979. From left: Contest Organizer Roger Allan, Art Boland and Ron Dilbeck. - 1979 NCC Finals (New York): The PNNL (Battelle) team swept the major categories: [10][5]
- Moonlight Express: An upgraded version of Moonlight Special featuring high-torque motors and finalized backtracking logic. It won the award for Fastest Learning Mouse, finishing in 31.16 seconds on the shortest path, on its third run. The team was awarded a Telequipment S61 oscilloscope donated by Tektronix.[5]
- Moonlight Flash: A high-speed wall-following variant with optical sensors that won Fastest Mouse overall and a $1,000 grand prize for fastest first run, finishing in 30.04 seconds.[5]
- Moonlight Special: Competed with finalized backtracking logic (finishing in 50.38 seconds on the shortest path, on its third run) to demonstrate the repeatability and stability of the project's generalized maze-solving software. In addition it displayed its cute fiberglass mouse-like body, often seen online in an iconic double-exposure image created by a Battelle photographer.[5][4]
Legacy
Moonlight Special is regarded as a pioneer in the history of Micromouse and autonomous mobile robotics as the first Micromouse to demonstrate maze-learning capability. It was described by contest organizers as "the smartest micromouse observed" during the competition's preliminary time trials. Its ability to start at an undisclosed maze entrance location and solve for the shortest path to an undisclosed goal location via Graph traversal and implementing a Depth First Search (DFS) approach within a self-contained Z80 architecture, remains a foundational example of embedded systems-based artificial intelligence, and a seminal milestone in self-contained simultaneous localization and mapping robotics. [11][12]
Uniquely, the original 'Amazing Micromouse' contests (1977–1979) utilized mazes with undisclosed entrance and exit points along the perimeter, requiring a generalized graph-traversal approach like Tarry's algorithm to explore and map the environment. In contrast, competitions beginning with the 1980 Euro Micro event in London moved to standardized starting locations and fixed-center goals (often protected by a 'moat' to defeat wall-huggers), which allowed later robots to utilize coordinate-based heuristics rather than the more general search logic pioneered by Moonlight Special.[1][13]
The capabilities of maze-solving Moonlight Special caught the attention of PNNL Energy Systems staff member Bruce Cone who was leading a USDOE-sponsored $3.5M “New Technology in Agriculture Program” in 1978. Bruce said, “If our staff members can design a mechanical mouse to solve a maze, why can’t we develop new ideas for farm machinery?” Thus, Moonlight Special moved from a computer contest entry only, to also a serious example of pioneering robot technology - that could help move farming into the age of automation.[14]
In June 1979, a Battelle publication featuring Moonlight Special reinforced the broader robotics potential of the project: “There are dozens of industrial and domestic applications which would free people from well-defined, monotonous tasks. The micromouse is an example of how microprocessors might be used. It’s a true robot, requiring technical and electro-optical technologies.”[15]