Tinypiper testing and further development
The third sprint consisted of water-testing boat designs, completing the path planner, and making additional headway on other projects. Being November, it is now definitely too cold for any testing on Lake Waban, and we are restricted to testing in the LPB and Babson pools from now up until the Great Thawing that is spring. But we are not deterred, and we continue to make progress despite the attempts of the natural environment to undermine our efforts.
The Mechanical Subteam finally, after many training delays, gained access to Olin’s Large Project Building (LPB — a small building for large projects). The LPB pool has been filled with water, meaning that Mechanical was able to test their Tinypiper prototypes. They created a small experimental rig that could drag each Tinypiper across the pool, which they to qualitatively analyze the hydrodynamic performance of each design. After testing five unique Tinypipers, the Tinypiper with the hull shape most similar to the current CAD assembly for Hawsepiper stuck out as the hull shape with the least drag.
Given those results, Mechanical will be moving forward with the current Hawsepiper hull design. All efforts will be focused on finishing the CAD for Hawsepiper and creating prototypes of Hawsepiper’s various parts. This includes creating a full-scale hull prototype and a wingsail prototype.
On the software front, the Planning and Controls finished implementing the path planning algorithm proposed by Roland Stelzer in his PhD thesis. The first run (video) wasn’t perfect but elicited celebration nonetheless. Fortunately, the bugs were easy to fix, and a successful virtual run was completed before the start of the sprint review (photo).
Additionally, the simulator now models the physics of the wingsail to determine the correct lift and drag for a particular wind speed and angle of attack. ROS integration is coming soon, at which point we will be able to test the path planner with actual physics simulation.
Furthermore, the computer vision group continued working through various resources on machine learning and computer vision, including the Machine Learning Crash Course by Google Developers. Their goal has been to build a base understanding of neural networks in order to better understand and use an object detection and classification neural network. (This could be Faster R-CNN, YOLO, Mask R-CNN, or another model.)
Lastly, we decided not to use the Airmar on Hawsepiper, due to reports that salt water builds up on its ultrasonic sensor membranes and interferes with wind speed readings. Instead, we are going to use an old-fashioned weather vane and an Adafruit bno055 IMU for orientation. Efforts to use the one already in our possession have been fruitless thus far, so a new unit was ordered in case the original had been damaged. The replacement arrived on the last day of the sprint, so testing will commence during the next sprint.
The Electrical Subteam has tested all of the parts needed to make Jankboat remote-controllable. With some more time and soldering, they are confident Jankboat should become fully remotely operable by the end of the next sprint. Electrical documentation is also coming along, and members are scheduled to get EE protoroom trained shortly.
We were relieved to find that our current Hawsepiper hull design is reasonably hydrodynamic and thus does not require revisiting. All in all, however, progress has been slowed slightly by the departure of a number of first-years from Mechanical. (This is to be expected during the first semester, when first years are encouraged to try out numerous clubs and project teams.) This has created a bit of a bottleneck with mechanical design and fabrication of Hawsepiper, especially since only one member is trained to do contour cuts on the ShopBot. While these facts will hinder our progress somewhat, we will rise to meet the challenge.