Robot Accuracy for Automated Inspection and Metrology Applications

The problem

Robot-mounted inspection — structured light scanning, laser profilometry, vision-based dimensional measurement, ultrasonic testing — uses the robot as a positioning platform for the sensor. The measurement quality depends on two things: the sensor's own accuracy and the accuracy with which the robot positions and moves the sensor through the measurement volume.

For contact inspection (ultrasonic, eddy current), the robot must place the probe at the correct position and orientation with respect to the workpiece surface. For non-contact inspection (3D scanning, laser profiling), the robot must traverse the scan path at a controlled speed and known position so that the point cloud can be accurately registered in the workpiece coordinate frame.

In both cases, the robot's positioning and path accuracy directly affect measurement quality. The error sources are the same as in any robot accuracy application — kinematic model errors, joint compliance, vibration, thermal drift — but the consequences are different. In a manufacturing process, a positioning error produces a defective part. In an inspection process, a positioning error produces a measurement error — which either falsely rejects a good part or falsely accepts a bad one.

Error sources specific to inspection

• Vibration during scanning: Robot-mounted sensors acquire data during motion. TCP vibration adds noise to the measurement — directly in the case of contact probes, and through registration error in the case of 3D scanners that rely on robot position for spatial referencing.

• Kinematic positioning error: If the scan path is programmed from CAD coordinates (offline programming), the gap between the nominal and actual robot kinematics creates a systematic offset between the intended and actual scan positions.

• Thermal drift during measurement campaigns: Inspection routines can run for hours. Accuracy drift during the measurement campaign introduces systematic error that varies with time, not with the part being measured.

• Path accuracy during continuous scanning: 3D scanning requires the sensor to follow a path at controlled speed and standoff distance. Path accuracy errors change the scan geometry and affect point cloud quality.

Compensation approaches

• Kinematic calibration corrects the geometric model, improving positional accuracy for programmed scan paths and reducing systematic registration error in point cloud data.

• Vibration compensation reduces TCP oscillation during motion, improving the signal-to-noise ratio for sensors that acquire data while the robot moves. This is particularly relevant for scanning applications where the robot traverses the measurement path at speed.

• Dynamic joint tracking compensates for joint-level errors during operation — backlash, compliance, thermal drift — providing continuous accuracy correction throughout long measurement campaigns.

Performance context

• Kinematic calibration can reduce absolute positioning error from ±0.5–1.5 mm to sub-millimeter levels

• Vibration compensation reduces dynamic TCP error during scanning by greater than 80%

• Combined approaches address both static (positional) and dynamic (vibration, drift) error sources

Integration

• Works through the robot's command interface — no controller firmware modification

• Compatible with any sensor payload; does not require sensor-specific integration

• Calibration is per-arm, capturing the specific unit's characteristics

• Can be applied to existing inspection cells without mechanical redesign