Control base class.
Applies no action (sets voltages to zero).
Applies a random action (samples from the action space).
Uses PID control to minimize (pitch - reference pitch) + (yaw - reference yaw), where reference is the original position.
Uses a mixed mode controller that uses gains found from LQR to do the flip up when the pendulum angle is over than 20 degrees off upright, and uses PID control and filtering to hold the pendulum upright when under 20 degrees.
Holding control uses PID with filtering, and outside of 20 degrees use no control.