We arrived in San Diego with a lot of work to do. We had months of preparation behind us, but water sealing issues had prevented us from testing many of the electronic systems we intended to use. Our first failed water test was in June, when the latches that were supposed to keep the main compartment sealed popped off underwater. After a redesign, we were confident that the main sealing system would work. However, we had a new problem. Running wires from inside a sealed hull to something outside is tricky. The wire needs to pass through without letting water in. Most teams use expensive metal connectors. Our cheap plastic Bulgin pass-through connectors, which had held through the first water test, began leaking in our later tests in Montreal. Making matters worse, while there were only 2 connectors in the first test, there were now six. So when we arrived at the competition, our hull was failing every water test with water pouring in at an alarming rate. We could not risk the main computer from Lippert of the IMU from MEMsense. Both were too valuable to be trusted to a leaky hull. This prevented us from attempting 
any parts of the competition except for the qualifying gate. The qualifying gate is a large square placed just under the surface of the water. If the robot can pass through it without hitting the any of the edges, it qualifies for the rest of the competition. The AUV is aimed directly at the gate by a diver, so the only requirement to pass through it is to be able to swim in a straight line. Some of the best teams don't bother using vision to pass the qualifying gate. They simply maintain a compass heading and rely on the diver's aim to send the AUV through the gate. We had no vision, no sonar, and no compass available so navigation was not a possibility. Our only hope was to make MARTy swim in a straight line by balancing the weight and thruster intensities.
any parts of the competition except for the qualifying gate. The qualifying gate is a large square placed just under the surface of the water. If the robot can pass through it without hitting the any of the edges, it qualifies for the rest of the competition. The AUV is aimed directly at the gate by a diver, so the only requirement to pass through it is to be able to swim in a straight line. Some of the best teams don't bother using vision to pass the qualifying gate. They simply maintain a compass heading and rely on the diver's aim to send the AUV through the gate. We had no vision, no sonar, and no compass available so navigation was not a possibility. Our only hope was to make MARTy swim in a straight line by balancing the weight and thruster intensities.In addition to the water-sealing problems, we had only two motor controllers at the time of the competition. Each controller can control two thrusters, so that limited us to 4 thrusters instead of the 6 in our ori
ginal design. It also meant that we had no backups in case any of the motor controllers broke during the competition.
ginal design. It also meant that we had no backups in case any of the motor controllers broke during the competition.Our first three days at the competition were devoted to last-minute construction. We slept very little and worked constantly while we were awake. We were finally ready to water test again after a few days. We had resealed all of the pass-through connectors to make sure they were attached properly, and the frame was ready with four thrusters mounted.
We had two Seabotix BTD150 thrusters, an expensive unit that performs very well. We didn't have enough money to buy any more, so we had to improvise the other two. We modified bilge pumps to work as cheap thrusters. In order to do this, we removed the bulky parts from the bilge pumps and replaced the impeller in each with a propeller by designing drive dogs and adaptor tubes. We soldered everything together, and the result was pretty impressive once we got it in the water.
With four working propellers, we had everything we needed to make a shot at the qualifying gate except for a sealed hull. We were willing to sacrifice all of the electronics inside the robot, but if the hull leaked too fast we would fry the electronics before we could even make it to the gate. The pass-through connectors were clearly not up to the task, so we took it upon ourselves to make them seal. One thing we packed plenty of was adhesive. We had an impressive array of waterproof sealants in our bags. We slathered every connector in layers of silicone, polyurethane and epoxy. Electrical splices were coated in electrical tape, slathered in epoxy and finished with a layer of silicone. When we were done, the leaking had slowed to a manageable pace.
We were finally able to test MARTy in the water with thrusters. Late in the third night of the competition, we ran our first tests with four thrusters in the pool at our hotel. We woke up the next morning convinced we had a good shot at the qualifying gate. That was when we lost our first motor controller.
The day before our first complete pool test, we had done some early tests in the kiddy pools next to the competition pool. We had noticed strange behavior and we didn't understand what was causing it. Every AUV has to have a kill switch. This is a button that the diver can press or pull or twist to make the robot stop moving. Ours was modeled after the system used by ETS, and reflected our team's Canadian origin. Magnetic switches allow a switch to be activated without any physical contact. If a magnetic sensor is placed on one side of a barrier, a magnet can be held up to it to activate it. We placed a magnetic reed switch against one of our endcaps, and glued a Canadian quarter to the other side. Canadian quarters are magnetic, so the magnet could be attached to the quarter to flip the switch inside the hull. The system 
worked well. During one of or practice runs, we noticed that our thrusters had unexpectedly run at full speed and would not stop, even when we pulled our kill switch. We tested the kill switch, which still worked, and were baffled by what had happened. The next day, the motor controller that had gone crazy did not work at all. It was covered in rust. The motor controller had gotten wet the day before, when it stopped responding to the kill switch, and had now deteriorated beyond working condition. We were down to two thrusters.
worked well. During one of or practice runs, we noticed that our thrusters had unexpectedly run at full speed and would not stop, even when we pulled our kill switch. We tested the kill switch, which still worked, and were baffled by what had happened. The next day, the motor controller that had gone crazy did not work at all. It was covered in rust. The motor controller had gotten wet the day before, when it stopped responding to the kill switch, and had now deteriorated beyond working condition. We were down to two thrusters.There was still a possibility of passing the gate. If we angled our AUV down, it could still follow a path through the gate. We stayed up late in the hotel pool tweaking and testing.
After a long night of testing, we gave it our best shot. Unfortunately we were not able to make it through the gate. Part of our attempt is shown in the video below.
Despite MARTy's poor performance, we had a great time at the competition. We learned a lot by talking to the other teams, and we know what we need to do to succeed next year. Most of our problems stemmed from the fact that the hull was not water-sealed. We learned a lot about effective water-sealing by looking at the other robots. Since we already have an IMU for navigation and a fast enough computer, we can start working on MARTy's ability to pass the course as soon as we have our new hull.
