This week was mostly focused on writing a prelab report and setting up the experiment for the data collection.
- Prelab : A prelab report for the experiment was written so that all the project team members and I could prepare for the professional RC driver to stop by and collect data.
- Optitrack Motion Capture System :
Thanks to Amir's insightful tip on calibrating the overexposure settings on the Optitrack system, we were able to finally localize the vehicles in a robust manner.
The key point was to calibrate the overexposure to a level which was sufficiently low enough to capture all the Optitrack markers. There were some trial and errors but eventually we empirically hit the sweet spot value which would allow robust localization.
- 3D Printed Parts :
An exhaustive amount of time was spent attempting to 3D print parts. The biggest problems were that the filament we used was too cheap to work robustly. Lesson of the day is to never cheap out on printer filament. It takes an exhaustive amount of time trying to declog the nozzle and take out the broken filament piece stuck inside the plastic extruder tube. Another problem was that the printer had a tendency to make the parts warp on the heated surface. The lesson learned here was to always use rafts when printing parts. The problem with rafts is that the finish is very dirty on the surface of the part which touches the raft. The key is to minimize this surface area to insure that the parts are still clean. Finally, the expensive filament from Ultimaker had trouble extruding evenly for small nozzles of diameter less than 0.4mm. In the end, I used a 0.8mm nozzle to print the parts.
- Noob Welcome :
Another portion of my time was spent covering the project idea to my new project members who recently joined the lab. Ryan will help out with the mechanical side of the project such as 3D printing, Camera distortion matrices and the Optitrack System. Kevin will be helping out by testing out the Neural Network FPGA board with the dataset collected from the prelab experiment as well as the beta testing phase(Where I drive the vehicle slowly and collect data) of the experiment. Cavan is focusing on the Unity Simulator development, which includes the construction of rigid body physics inside the simulator. He is using the point cloud scan I rendered within the Dot 3D software using the realsense camera.