Mounting, Calibrating, and Trusting Your IMU

An IMU is only as good as how it is mounted and configured. A perfect gyro on a vibrating, loose bracket will give you garbage. This lesson is the practical checklist.

Mounting#

• Mount it flat and rigid. The device should be parallel to the floor and bolted to a solid part of the frame, not zip-tied to a wire bundle. Vibration and flex add noise that integrates into drift.
• Keep it near the center of rotation when practical and away from large vibration sources (like compressor pumps) and strong magnetic fields (the magnetometer in 9-DOF units can be disturbed by motors and steel).
• Note the orientation. If the chip is mounted sideways or upside down, the axis you read for yaw changes. The Pigeon 2.0 lets you store a mount-pose orientation; otherwise account for it in code.

Calibration — know your device#

Calibration steps depend on the specific IMU, so read its docs:

• Boot/bias calibration: Some IMUs (for example the older NavX/NavX2 and the original Pigeon) sample their zero-rate bias at startup; for those, hold the robot completely still during boot, or the bias is wrong and drift is worse. WPILib's analog gyros expose a calibrate() step done on a stable surface. The Pigeon 2.0 is different: it requires no boot or temperature calibration and does not need to be still at startup.
• Mount calibration (Pigeon 2.0): Run the one-time mount calibration in Phoenix Tuner X after the device's placement is finalized so its axes line up with the robot.
• Level/offset setup: For devices that support it, set offsets so pitch, roll, and yaw read zero when the robot sits flat.

Zeroing heading#

There is a difference between the gyro's internal calibration and your field heading. At the start of a match you typically reset/zero the heading so the robot's current facing maps to a known field angle (e.g., set the pose/heading from your autonomous starting position). Provide a driver button to re-zero in case of a mid-match disturbance, but be careful: re-zeroing during teleop can confuse field-oriented drive.

Validate before you trust it#

Never assume the sign or scale is right. Quick checks:

• Spin the robot exactly 360 degrees by hand and confirm the reported yaw changes by ~360, in the CCW-positive direction WPILib expects (invert if not).
• Let the robot sit still for 30-60 seconds and watch the heading on telemetry — a good modern IMU drifts only a fraction of a degree; large drift means a mounting or configuration problem.
• Drive a straight line and confirm field-oriented controls feel correct.

These five minutes of validation prevent the classic 'the robot drives sideways in autonomous' disaster.