Young teams dominate this year's Mobot race
Wheels, or sometimes feet, propel them; skirts of trash bags hang from their sides. They move of their own accord, making the hours of work that go into their production worthwhile — or tragically wasted. They are mobots (mo*bile ro*bots), and every year crowds come to watch new robotic vehicles cooked up by the students and affiliates of Carnegie Mellon.
This year, the entries were sparse. Ben Brown, a member of the robotics faculty and the Mobot Committee, noted that in past years there had been as many as 20 entrants. Among the undergraduate, open, and exhibition classes, there were only 10 entrants, and of these, three teams never showed. The undergraduate class awards cash prizes for first, second, and third place. The open class is for members of the campus community who are not undergraduates, and it awards a single prize for best run. The exhibition class is simply for those who wish to show off their mobots, and not to compete; therefore, there is no money involved in the exhibition class.
The Mobot course along the sidewalk in front of Wean Hall is a treacherous thing, like a tight turn on a slick road. The mobots must be programmed at their most basic level to recognize a line. However, the cracks in the sidewalk and the slight differences in the sidewalk color make this a more challenging task.
In addition, the mobots must deal with two hills: one just after Gate 1 and one just after Gate 8. The second hill is much steeper, but both hills are difficult challenges because mobots can easily lose control, flip over, or lose sight of the line.
There is a decision portion of the course where mobots must decide whether to turn or not, based on the positions of the gates. At this point, the course changes yearly; sometimes the first decision will be to go left and sometimes to go straight. Teams must be prepared to change their robot’s program to fit the way the course is laid out in any given year.
On account of these multiple challenges, most teams do not make it to the decision points of the course, or, if they do, they do not have an effective solution for these points.
The big winner in the undergraduate class was Pajamas, which resembled a remote-control vehicle with its guts exposed. Its run was impressive — it gave the audience a scare as it stopped on the cracks for a moment but was then propelled forward with a jolt. Joey Gannon, a co-creator of Pajamas, got harassed a little for following the car with a video recorder — some onlookers thought he was operating the mobot via remote control. Pajamas skillfully dealt with the hills by continuously jerking its front wheels slightly right and left to create resistance and make it slow its pace. Its major bout of trouble came after the second hill, during the part of the course where the vehicles must “choose” which of the white lines to follow. Pajamas stalled out after Gate 9, apparently unable to decide.
Gannon and Daniel Dewey, the creators of Pajamas and both first-year computer science majors, started working on the mobot in October. Their greatest challenge: “The cracks in the sidewalks,” said Dewey. One thing that Pajamas can do is make decisions and know how to behave based on the number of gates it’s crossed. This is how Pajamas knows when it is at the point of the course where it must make a choice. Another challenge may be solved in future years with a little collaboration. “We would have appreciated a MechE,” said Dewey, noting that the task is mostly a mechanical one rather than a programming challenge.
Some undergraduate teams were less fortunate. Mission Possible, a rather small tri-wheeled vehicle with a collection of wires poking out of its top, had the first run of the day. It got about a foot, then veered off the course in a disappointing foul. On its second run, Mission Possible again stalled no more than five feet from gate 0 (the starting gate). Del Salvador, another undergraduate entry, was unfortunate in that it took a wide right about 10 feet into the course. Del Salvador also resembled a remote-control vehicle and moved with breakneck speed. But during both of its runs it seemed to forget to follow the line.
The open class was sparse pickings and ended up being dominated by a competitor who knows the race very well. John Palmisano’s mobot Pikachu won the open class prize, becoming the third mobot to earn Palmisano an award. An alumnus from the mechanical engineering department, Palmisano won the judges’ choice award last year and third place two years ago when he was a senior. Pikachu couldn’t blast through the course on its first run, though. The mobot stalled out and freaked after a hard right, but it came back big in the second run. After minor adaptations to the trash bag skirting its edges, Pikachu scurried off along the line. It quickly made it to gate 8 before ignoring some of the vital gates in the decision portion of the course.
A particularly interesting addition to the open class was the group The Young Ones, led by a PhD student in psychology and his two young daughters. This diminutive mobot didn’t get far, but it was great to look at. It had a tiny front wheel and two bigger back wheels. On its second run, the younger daughter, Victoria Douglass, placed a small stuffed cat on top of the vehicle. Though it never made it to Gate 1, it was charming in both its size and additional decor.
In the exhibition class were some old favorites and a new addition that was a crowd favorite. One of the few mobots not strongly resembling a motorized car, The Little Dog was actually the most expensive participant ever entered in the race. The body is a quadruped robot named LittleDog and manufactured by Boston Dynamics. Its operator, Michael Dille, made a few adaptations so that it could run the mobot course. Visible additions to bring The Little Dog to mobot level were some taping of the paws and the addition of a massive camera on its back.
The Little Dog had some issues, however. First of all, it moved at a sluggish pace. And second, the robot seemed attracted to the white dots that sometimes frame the line along the mobot course. The Little Dog only made it through Gate 1 with a little nudging to get him back on the line. And he crashed going down the first hill, toppling over and knocking his camera into pieces.
Though only a few mobots got past Gate 1 this year, it’s worth noting that the challenge is by no means an easy one. As Brown said, “It was the eighth year [of Mobot] before anybody [finished the course].” Over the 12 years that Brown has been judging, the main change he noted was the shift to video cameras over other varieties of sensors.
Behind the Mobots
Later, at the 15th Gate Ceremony (so named because it follows the races and there are 14 gates on the Mobot course), Brown allowed the three teams who earned prizes to come up and discuss their creations. Gannon and Dewey explained how Pajamas made it through the decision points using its sensors. The robot is told that its first priority is to avoid putting its rightmost or leftmost sensor on the line, and in its zeal to comply it causes the mobot to turn.
During its runs, Pajamas seemed to stall out for a moment on the cracks in the sidewalk. Dewey explained how the mobot kicked itself into gear again. “We have a function called ‘Don’t get stuck,’ ” he said. “If the sensor information is the same ... it revs up the motor.” In other words, if the mobot is staying still, the sensor can tell it’s not getting new information and the engine gives it the extra boost to make it past the cracks.
Next, Pikachu’s “master” Palmisano told of his very different idea about how to build a mobot. He used a simple design and chose to go with photoresistors that pick up green light, which he claims the white Mobot line reflects more than other colors. Pikachu was built in three weeks, said Palmisano. Explaining the choice of photoresistors over an infrared emitter detector, he said, “I was lazy.”
Finally, the judges’ choice winners, The Young Ones — Scott Douglass and his daughters Victoria and Rebecca — spoke about their mobot. Douglass’ daughters have been watching the races for years, he said. “It’s ... better when you get to participate,” the youngest Douglass noted. Scott Douglass also explained that this mobot was an adapted LEGO kit, which might explain its small stature.
Whether big or small, however, mobots add a certain flair to Carnival. Amongst the parties, booths, and shows of athletic prowess like Buggy, the mobots remind us of the things we are most often recognized for. Carnegie Mellon: technically minded, and proud to show it.