My name is Peter and I am a Computer Engineering student at Santa Clara University. I am working on a team of three to build an e-Foil and related software products for our senior design project. We will be presenting our work on May 28th (along with 70 other diverse engineering teams) via YouTube live stream at the Santa Clara University annual Senior Design Conference. The conference is being held online for the first time due to covid-19 for the public to watch so I encourage all to watch and see what entry level engineers at the Jesuit school in Silicon Valley are working on.
So you’re building an e-Foil? Sort of.
Like many here, I enjoy kiteboarding. My go to spots are 3rd Ave and Sherman Island. I wanted an e-Foil for obvious reasons. To finish my Computer Engineering degree I needed a project and building an e-Foil falls more into Electrical Engineering so to make it work I would need to have a strong software component. I read efoil.builders for months and recognized few users used data to refine their designs. Metr.at was a step in the right direction but I was turned off by the fact they had made their own expensive proprietary transmitter. We would not have the VESC where it is today if Vedder had not made it free and open source; for Metr.at to make a closed add-on goes against the spirit of open source.
Using a generic BLE VESC6 module and the $2 Vesc Tool Android app (supports Vedder) I could log my own data! Metr.at still wins over this approach because of it’s great plotting capability. Logically it follows, how to plot the Vesc Tool data in a great way? With my own software!
Vesc Tool Plotter is the software component of the project. It is a web app (in progress) where you can upload a CSV generated by Vedder’s Android app and get a sharable link to a plot of your ride similar to Metr.at. You will have the ability to specify different parameters about your build like which battery, motor and foil you are using. You can search for data from other builders based on these parameters. The goal is to help humans analyze different builds to refine open source e-Foil design.
The scope of the project for our presentation ends there. We were building an e-Foil prior to covid-19 to collect data for the app using designs from efoil.builders however it is now considered extra. I am still building it of course. I will be posting in here over the next month an accelerated e-Foil build where I take the KISS principle into to uncharted waters. With 70% of the parts on hand and 29% in the mail we will see if it is really extra or not.
A final word of the data driven e-Foil design approach. As I mentioned Vesc Tool Plotter is designed for human analysis but the underlying architecture is not. By recording large amounts of labeled data my stretch goal is to analyze e-Foil designs with machine learning algorithms. e-Foils are complex systems which are governed by the laws of fluid dynamics other complex physics. I have no clue how each component of our system interacts but I do know how to train a model which does. A benefit to this approach is by analyzing the whole system (as opposed to discrete pieces) the relationships between components are present in the data. My pipe dream is to have a model trained with a reward function which minimizes current draw using features like propeller pitch and diameter. I could go for rides on several different 3D printed propellers and view the plots of the effects against each other using Vesc Tool Plotter then get a recommendation from the ML algorithm to, for example, decrease propeller pitch to improve my efficiency. Now that’s data-driven e-Foils!