2017 Day 2

Written by James Irwin. Posted in Competition

This morning we finished the wiring for charging our new battery within its case, which saves us much time and reduces the probability of damaging seals or wires. However, we did realize that our cable placement makes it difficult to charge the battery while it is mounted on the sub, which is an issue we'll definitely be fixing next year.

We calibrated our new PNI TRAX IMU, which significantly helped with our yaw drift. The drift still exists, but it is much more manageable and we don't expect it to be an issue for the tasks we want to complete at the competition this year.

We performed more testing of our gate task and it did not perform as well as it had been in simulation. We largely attribute this to mistakes by the vision processing system, and also the fact that the PID parameters of our control system require more tuning. This evening we performed some PID tuning until our battery was exhausted and plan to complete the tuning tomorrow.

About half our members worked on labeling more vision data today, while the rest worked on the physical sub on sensors, hardware, and AI tasks.

 

 

We are training our neural network overnight on the newly labeled data, so vision processing should perform much better tomorrow!

2017 Day 1

Written by James Irwin. Posted in Competition

Today was a great first day of competition! While we were a bit late waking up in the morning (about 30 minutes), we quickly made up that time by qualifying during our first practice run in the morning. We had initially planned on doing only data collection during the first practice run, but decided to see if we could qualify. After some tethered work to ensure our control system and sensors were working properly, we attempted to "bowl" the gate, meaning that we told the sub to hold a constant depth and yaw, and to blindly travel forward. We are proud to say that we passed through the gate and qualified on our very first attempt, which I believe is a first for our team! We also were able to collect a huge amount of vision data that we are labeling tonight to continue training our neural network for object detection.

The new battery case has been completed and confirmed waterproof. It was successfully used during our second practice run, which we used for collecting more vision data.

A few members have been working on getting hydrophones working and have made good progress today. We hope to have them working stably off the sub tomorrow, and are considering integrating them into the main sub. This would allow us to attempt much more complex tasks and potentially enter finals.

Relocating our IMUs has helped significantly with the yaw drift issues we have been seeing before the competition, however we are still working on reducing the drift by calibrating our sensors.

We also discovered various tweaks that we needed to make to our software, including scripts to save time while doing runs and adding support for reading the temperature sensors in our computer and IMUs to ensure things don't get too hot when the sub is on the shore. 

Our new router box has been assembled and verified working, and it has made a significant difference in our network performance. We now have greatly reduced lag when streaming live feeds from our cameras.

Unfortunately, we did have a hiccup at the end of our day while leaving the TRANSDEC facility. On the way back to the hotel we got a flat tire and needed to swap to our spare. Early tomorrow morning we will be getting the tire repaired at a tire store. This was only a small bump in what was otherwise a very promising first day!

Tonight we are labeling data, calibrating sensors, and making small tweaks to our software. Our goals tomorrow are to reliably travel through the start gate via vision feedback and to hit the buoys.

Competition So Far

Written by Daylan Kelting. Posted in Competition

Upon arriving at the Transdec facility in San Diego we dove straight into our work. The sub was in a few pieces and needed to be completed before our first runs could begin. We worked hard and late into the night to qualify on time. 

By Wednesday we had assembled our sub. For the first time, the sub made a foray into the depths of a testing pool and, to our elation, it was buoyant and rather stable.

However, our good fortune did not last much longer. Upon our return to our hotel, five of the eight thrusters failed as soon as we started testing. After much deliberation and much troubleshooting, we made the difficult decision to resort to the previous year’s sub and move forward with qualifying.

Some fiddling later, the back-up vessel was in the water. On Thursday, we were able to start doing runs to qualify. In the first run, we discovered some kinks that we had to work out. Our second run was more successful and resulted in our qualification into the semi-finals.

However, more issues arose and troubleshooting was a frequent occurrence. Just today we adjusted our priorities, deciding to focus on learning as much as possible from other teams rather than spending tons of effort getting the old sub to perform even as well as last year. The next few days will be spent learning as many lessons as we can from other competing teams and improving for next year.

Bookmaker review site http://wbetting.co.uk/ for William Hill