Characterizing a Mechanism with SysId

Rather than guess feedforward constants, you can measure them. WPILib's SysId (System Identification) tool runs your mechanism through controlled tests, records data, and computes kS, kV, kA, (kG for arms/elevators), and starting PID gains.

How SysId works#

SysId runs four tests, each forward and backward (reverse):

• Quasistatic: the mechanism is sped up very slowly so acceleration is negligible — this isolates kS and kV.
• Dynamic: a constant 'step' voltage is applied so the mechanism accelerates — this reveals kA.

During each test SysId logs voltage, position, and velocity. It then fits these to the feedforward model to extract the constants. Because the fit uses real data from your robot, the numbers reflect your actual friction, gearing, and mass.

The modern workflow#

In current WPILib you create a SysIdRoutine in your subsystem. You give it a way to apply a voltage and a logging callback that records the motor's voltage, position, and velocity, then bind the four test commands (quasistatic-forward/reverse, dynamic-forward/reverse) to buttons:

SysIdRoutine routine = new SysIdRoutine(
  new SysIdRoutine.Config(),
  new SysIdRoutine.Mechanism(
    voltage -> motor.setVoltage(voltage.in(Volts)),
    log -> log.motor("arm")
             .voltage(appliedVolts)
             .angularPosition(positionRad)
             .angularVelocity(velocityRadPerSec),
    this));

The Config lets you set the quasistatic ramp rate, the dynamic step voltage, and a timeout so a mechanism does not run into its hard stop. CTRE's Phoenix 6 integrates with SysId as well, logging via signals for Talon FX systems.

Safety and good practice#

• Give the mechanism room to run, and set a short timeout; for limited-travel arms/elevators use a small ramp/step so it does not slam a hard stop.
• Run each test and stop before the mechanism reaches its limit.
• Load the resulting log into the SysId analyzer, pick the correct mechanism type (Simple, Arm, or Elevator) so kG is computed correctly, and confirm the fit looks clean (a good R-squared).
• Treat the PID gains SysId suggests as a starting point, then fine-tune with the methodology from the earlier lesson.

SysId turns control tuning from guesswork into measurement, which is the difference between a robot that occasionally works and one that repeats the same shot all match.