6.270 is an annual student run robotics competition held at MIT each January. Teams of two or three students are given one month, a big box of Legos, a controller board, a band of hard-working organizers and TA's, and told to build a working autonomous robot. At the end of the month, the teams compete their robots in a two day round-robin tournament. This year's competition was Bots in Blue.
Who IS Uncle Seymour?
(From the Oxford English Dictionary)
uncle ('/\jk(&schwa.)l), sb. Forms: a. 3-7 vncle (5 wncle), 4- uncle
(5-6 oncle). B. 4, 6-8 unkle (vn-) 6-7 unckle (vn, 7 wn-). y. 4 unkel,
5 vnkel, 6 unkell (vn-), 5-6 vn-, unkil(l, -kyll (5 hunckyl, oncyll,
ownkyll, 6 onkill)
aC. 1290 S. Eng. Leg. I. 20/27 To his vncle he gan go, Pe Erchebischop
of caunterburi. 1297 R. GLOUC. (Rolls) 1937 Pre vnclen is moder adde. C.
1380 Sir Ferumb. 169 Pe duk..drow ys swerd anon, & wolde ys vncle par-
wip herte. 1475 Bk.
Noblesse (Roxb.) 15 Henry the v[te]..made Thomas Beauford then erle Dor-
set hys oncle capteyn of yt. C. 1500 Melusine xix. 97 For neuer I shall
ete tyl that ye be hanged with your vncle. C. 1520 BARCLAY Jugurth vi.
(1557) 6 b, Se that ye worshipe and loue this Jugurth your worthy vncle.
(from the Kabalarian Philosophy website--teaching the principles of Mental Freedom)
Seymour: The name of Seymour creates a very aggressive and independent nature, one with big ambitions, giving you original, progressive, large-scale ideas, salesmanship and promotional ability...You have a versatile, restless nature, and could do any job well, although you would not like to do menial tasks...Your intense, restless nature can bring on tension which affects the solar plexus and digestion, or the generative organs.
(Checking out the opposition on competition night. Dan doing a bad job of hiding his strange itch....)
We designed our robot for speed and ruggedness. We managed to balance these two seemingly contradictory goals well, creating an extremely strong robot that was also the second fastest out there (the only team that could possibly out-race us was Team 59, My Little Pony).
Drive Mechanism/Chassis
Our drive was built with the sole goal of maximizing speed. Our robot independently drove each of the front two wheels using three bi-directional motors per wheel. Because we needed small wheels to give us a larger wheelbase, we used a motor to wheel gear ratio of 5:1, probably the lowest gear ratio this year (typical values ranged from 27:1 to 45:1). Turning was accomplished by differentially driving the front wheels and counting the distance using shaft encoders, giving us fairly accurate turns.
Unfortunately, our chassis shape caused some problems. Firstly, we designed the chassis to hang over the first two steps of the center hill, taking away a few inches from our wheelbase and hurting stability. We saw this everytime the robot slammed into a wall and tipped forward. In addition, the ultra-low gear ratio made it difficult for our motors to precisely control our turns, because of the large amount of play inherent in the design. Despite these difficulties, the robot performed fairly predictably.
Arm Mechanism
Our front arms were designed to sweep the professors off the center hill without us having to climb any steps. We used long arms which folded into the robot in the starting position, each being driven by one bi-directional and one unidirectional motor.
Sensing/Feedback
We minimized the amount of feedback to maintain simplicity. Shaft encoders on our wheel motor gear trains helped us make accurate straight and turning motions. Touch sensors mounted on a front bumper under the arms let us know when we hit the hill in Mass Ave. Rear touch sensors allowed us to align with the rear wall.
We were given a controller board running NetBSD and an accompanying Java Virtual Machine. SIX DAYS BEFORE THE COMPETITION DATE. Because of the simple and sequential nature of our strategy, our final code was short and straightforward . We essentially ran an open-loop sequence with a small amount of feedback (touch sensors, a rudimentary line-following routine) thrown in. Due to the open-loop nature of the code, predictable operation of Uncle Seymour depended heavily on calibration of specific numerical parameters and the precision of the mechanical structure. We tried (and mostly succeeded in) meeting these requirements by robust mechanical design and by calibrating numerical values at the contest table.
(Calibrating right before round three. Note Andy Chang's exasperation....)