Summer recap

The team went to RoboBoat and RoboSub this summer.

At RoboBoat the team did well during the week. The first few days were spent collecting dynamics data to train the dynamics model and state estimation, and trying to get our new GPSs working. With the end of the week approaching, we finally resorted to using the same GPS as last year (Microstrain), and we got a good state estimation and controller for our boat. We were happy with our run in the semi finals: consistently doing the navigation gate, speed challenge, entering ‘Find the path’ and docking once.  Alas, Murphy’s law strikes again! And on the day of the final, due to an unfortunate GPS failure (busted connector), the boat got lost in the water in the middle of ‘Find the path’. The team won third place and is determined to come back next year!

At RoboSub the team did much better than last year. We focused on the first four tasks (starting gate, buoys, channel, pinger localization). Unfortunately the pinger localization system that was working in the pool was unable to identify the pinger at Transdec, and will be redesigned for next year! The first days were spent fixing some bugs and collecting data to test the computer vision. We qualified at the end of the first day. Vision algorithms were developed on site and proved to be robust, picking up the gates and channel very reliably. The buoys algorithm was still having some issue with detecting the green but could identify the red and yellow with success. Our best run was in the semi-final where we passed the gate, touched one buoy did a 720 spin in the channel and surfaced an inch away from the octagon. We qualified for the finals and were the first to go on a very sunny Sunday. Some condensation on the camera lens, as well as sun reflection, prevented the vision algorithms to work as well as they had before, and the vehicle did not go past the buoys. Overall, we were happy with our performance and we got fifth place in the competition, as well as the Teledyne Marine Innovation Award for our vehicle design.

As a conclusion, many lessons were learned.
Having a status display of the vehicle state (booting, ready, killed, recording…) is incredibly helpful to understand what might be going wrong, and this should be a requirement on any robot.
Focusing on core tasks is a good strategy. When trying to address every task the team always spreads too thin and usually ends up with every task only halfway done.
Designing for mechanical stability is far better than adding yet another actuator. It reduces the number of failure points and the amount of tuning required.
Having a stable platform makes the competition more enjoyable. Plan with the objective not to do any hardware changes at the competition. They are easy to make in the lab very hard to do in a hotel room, whereas in water time is hard to get in the lab and easy at the competition.

We’re excited to go back next year!