Day 19: Prototyping, Drivebase, and Programming
The shooter design is almost done. We are moving forward with 20 degree trajectory and tangent hood section out to 45 degree, to match the prototype. It should be ready to make this weekend.
The chute design is coming along. We will need to make it angled to fit gear grabber.
The gear grabber concept is complete and detailed design is moving along well. We should be able to make it by early next week. We need the chute mods to be able to hold the gear and fit it in the sizing box.
We worked on getting the superstructure ready for powder coating after welding. Then, we put holes in the frame to attach the superstructure to. We also then made some parts for intake well we CNC the other parts for the intake to get it ready for welding today. We also organized the metal in the shipping container.
On Tuesday, we tested our autonomous path code, which uses a web interface to plot an autonomous robot path. It worked; the robot moved along the plotted path with ~2 inches of its target! There were a few kinks that we are still resolving, though. The robot uses a NavX IMU (inertial measurement unit) to determine the robot’s heading, and the NavX board was throwing errors because there were multiple instances of the same board. This was resolved, but we still have to clean up the code to avoid these issues in the future. At the end of the path, the robot unexpectedly darted backwards; we have to fix that.. Finally, we tuned the drivebase PID constants for low gear but need to do the same for high gear, which the robot will likely operate in.
Today, we also revised the PixyCam software in order to make it more threadsafe. Rather than our current model, where we make calls to a PixyCam object, we changed the code to use a Listener interface- the PIxyCam object publishes an “event” whenever targets (aka “blocks”) are detected, on a different thread, and a listener on the main thread will handle each block when necessary. This makes the code more threadsafe- if the SPI driver crashes, it will not bring down the entire robot.
We also corrected the PixyCam distance estimation formulas to be much more accurate. Before, we were using the block height and width to calculate distance. However, the block height and width are very susceptible to noise and change minimally with distance. We switched to using the centroid of the target, which is far more evident to the PixyCam and therefore more accurate. We were able to calculate the distance to the boiler, plus-minus 10 inches, likely because of our setup and camera calibration.
For the drive code, we added in the ability to shift between low and high gear (using the pneumatics). We use the low gear mode on the robot when we need to move slower, but with more torque/power (e.g. when we’re aiming to shoot), and we use the high gear when we are barreling full speed down the field. Shifting is accomplished with pneumatics and dog collars. This is not quite a six-speed manual transmission, but two speeds is sufficient for our needs.