Testing Git Boat and Moving Forward
The second sprint is done, and the team has begun sailing along!
We attended the Boston Mini Maker Faire on October 6. We had a booth where we showed off Git Boat to visitors, and we met some OARS alums.
We also went to Lake Waban to test our retrofitted Git Boat. The winds were around 5–10 mph, and the boat tilted significantly under those conditions (below). This is very similar to what happened in 2014 (second below), implying our changes to Git Boat were not very effective.
We’ve thus decided that working on the Git Boat is not worth the time or resources required. From this point forward, we will dedicate our attention solely to Hawsepiper.
The Mechanical Subteam completed their Tinypiper CAD designs, and they are ready to fabricate them on the ShopBot. The plan for the rest of the month is to continue working on making and testing the Tinypiper designs, either in the LPB pool or in a “kiddie pool.”
Trainings to use the large project building (LPB) space and the ShopBot are still going slowly due to multiple delays on the part of the trainers. Thus, only three members are trained on the ShopBot, and lots of rescheduling still needs to happen. The LPB pool isn’t filled yet, but once that does open, trainings for that will be scheduled as well.
The electrical team has been moving along getting new members up to speed with PCB design and compiling all of their sensor libraries into a GitHub repository. They also gained another new member!
Within the software subteam, the path planning group has been working on developing an implementation of a short course routing algorithm based on the work by Roland Stelzer at the Centre for Computational Intelligence at De Montfort University. It works by incorporating the boat position, target position, and wind angle to determine the heading required to reach the target the fastest. The current implementation limits tacking, and can be adjusted to sail within a channel of set width. It is a better and more realistic implementation than the previous routing algorithm, but it still needs to be tested in a path simulation.
On the controls side, a team member is underway constructing a simulator for the boat, which will eventually integrated into a ROS node. With some hijinks here and there regarding the mechanics of sailing, the simulator is coming along but is still in the works. Some other team members worked on writing a ROS node for filtering the readings from the wind speed sensor, and then later worked on a learning exercise to get a better understanding of OpenCV.
A new group has also been formed to learn about machine learning and computer vision. To get started, they’ve been working through tutorials and other online learning resources. The eventual goal of the group is to implement either Faster R-CNN, Yolo or Mask R-CNN to recognize obstacles in images, which will trained and evaluated on water footage.
Lastly, some members have been working with the Adafruit bno055 IMU in search of more frequent and reliable heading readings, as the Airmar doesn’t publish heading information regularly. Unfortunately, neither previous code nor stock sample code seems to be working. An investigation as to the cause is still underway.
We’ve begun taking our strides into the water, but there are still areas to be improved upon. As a team, we’ve realized that we need to work on our adhesion to our central planning tool Asana. Thus, our kaizen for this coming sprint is “Asana.” A worry in our subteams is a slowing of our growth speed, but as we continue to review both what changes need to be made from last year and what must be started anew, we will hopefully only increase in momentum. And of course, there will be quite a few difficult waves ahead, but we’re ready to sail right through them.