Adaptable Flywheel Control

This robot did rapid fire meaning it shot 3 balls back to back extremely quickly. This meant the robots flywheel had to be able to maintain speed in this short time frame to prevent missing the ladder shots. To do this I first employed a simple PID Controller but I realized this either was too slow or too aggressive. Then I combined it with Feedforward so that feedforward would maintain the correct speed (more stable) and the PID Controller would react to the changes in flywheel speed. This worked better but I noticed the robot would, especially when not at full battery, undershoot the ladder balls. To fix this I made the PID part of the control system switch to a bang bang controller while shooting AND added Battery Voltage Normalization to the PIDF Control System.

Kalman Filter Real Time Localization

This year there were many benefits to being localized the entire time including turret automatic aiming, shooting while moving, autonomous movement, and determining flywheel velocity based on distance to goal. However, pure odometry and IMU readings are prone to drift and therefore as the round goes on, the robots position and orientation can become inaccurate. To fix this, I used April Tag Localization to get a global reference for the robot's position and orientation. However, this was prone to noise, especially at greater distances, poor angles, and while the flywheel was spinning.

To combine the approaches and reduce the cons of both, I used a Kalman Filter to smooth the noisy April Tag readings and estimate the robot's position and orientation over time, where odometry variance was proportional to the displacement of the robot, and April Tag variance was proportional to the amount of pixels of the tag (far away or poor angle = less pixels).

Automations

Turret Aiming

Given the robots current position, heading, and a goal position, the target angle of the turret can be calculated to aim at the goal. I did this by using trigonometry (atan2) to calculate the angle between the robot's current position and the goal position and then using the calculated angle and the robot's heading to calculate the target angle of the turret.

Dynamic Flywheel Velocity

To calcualte the target velocity of the flywheel, I created a manually measured regression line of over 20 data points that mapped the distance (X-axis, Independent Variable) of the robot from the goal to the target velocity (Y-Axis, Dependent Variable). Using this regression line, I could calculate the target velocity of the flywheel based on the distance of the robot from the goal.

Shooting While Moving

Finally, probably the robot's most impressive feat, shooting while moving, uses the following logic to generate a phantom goal for the robot to aim at while it is moving:

Above shows the robot shooting while moving.

Autonomous

To achieve an 18 ball autonomous, we used the robots many automations for shooting combined with Pedro Pathing's Pathing library for moving.

To collaborate with teamates who may be using the near zone, I created a far zone autonomous that would use opencv to detect balls and then have the robot use a PID Controller on the Drivetrain to move the robot to the largest cluster of visible balls.

TeleOp

This was the first year our robot won a FTC competiton since 2017. During teleOp the robot would implement the same automations as autonomous, but with a human driver controlling it.