Here is how competing students summarized the event.

The autonomous snowplow competition (ASC) is a competition run by the Institute of Navigation, with the aim of challenging students in the areas of robotic navigation and guidance. Student teams had to construct robot snowplows with the task of clearing an artificial test course. Teams scored points prior to the actual competition through preliminary design presentations and a team poster. At the actual competition, teams took turns in a best of two format and attempted to plow the competition field autonomously, avoiding obstacles, remaining in bounds, and returning to a designated garage area once complete. Teams are scored based on the number of obstacles hit, amount of snow cleared, and ability to return to the designated area.

Our team this year was a number of PhD students from the Biorobots group. Ian Adams, Shane Riddle, and Clayton Jackson under Dr Quinn, and Austin Mills, and Nathan Mengers under Dr. Daltorio. Our robot is Otto, a robot platform with a long history stretching back in some form or the other to Dexter and the DARPA urban challenge. 

We tend to use the competition as an opportunity to practice practical robotics, and to give students a chance to develop skills for all facets of robotics. Typically, students have a niche specialty in coding, manufacturing, or design, for instance. But as a team, we feel it’s very important to all be involved in all aspects. We arrived a day or two before the competition, to give ourselves ample time to correct any problems our robot might have had. After a few mechanical tunings, we were together as expected, collaboratively coding on the TV of our Airbnb.

We had the misfortune of randomly being assigned the first slot in the competition's morning trial. Naturally this meant that we would be pulling a near all-nighter to test our robot into the ground. After a few middling practice runs, just before 9PM we encountered an issue which would plague our runs for the next day: a bad circuit breaker. The robot expects to pull a large amount of current when operating, and even more under a heavy snow load. In this case, our circuit breaker was determined to be both the incorrect type, and also actively breaking down. This meant that the robot couldn’t perform even moderately difficult motion without tripping the hair trigger circuit breaker. We devised a way to prevent it from triggering, and turned in for a brief 2 hr nap, before having to get ready for our run. I wish I could say the first run went well, but it went as expected. The robot maneuvered into some snow, before repeatedly tripping the breaker and needing full restarts. Our score for the first attempt was only good enough for second to last place. 

We scoured auto parts stores, to dig up a suitable replacement for the circuit breaker, and rapidly set about replacing it. An advantage of having failed first, was we had more time to recover before the afternoon attempt. A part swap later, and we’re resuming testing. Soon enough a new problem arises: our navigation system worked sporadically. In certain conditions it performed fine, but in others it would seize and do absolutely nothing. In a rush, we decide to rewrite it, opting for a new take on older code from last year, rather than trying to polish the scratches out of our current code. We had just enough time to attempt a practice run before we were needed back for our last attempt. Our last test run wasn’t great but our time was up. Shane and Clayton manually edited some of the waypoints during the walk-up to the field, but now we had to run for real. No way to test our fixes first. 

And so, we pushed start. And the robot worked better than we had ever hoped. The only points where we were concerned were the turns around the obstacle, which thanks to the last-minute tweaks, worked perfectly. The robot neither hit obstacles nor left the competition field, and ultimately returned to the garage zone for a pretty perfect run, as much to the shock of the judges as ourselves.

Alignment

None