The Difference Between Accuracy and Repeatability

Before diving into diagnostics, it helps to be precise about terminology. Accuracy is how close a robot moves to a commanded position in absolute space. Repeatability is how consistently it returns to the same taught point. A robot can have excellent repeatability—within ±0.02 mm on a six-axis articulated arm—while still having poor absolute accuracy if its kinematic model does not match reality.

Most industrial applications rely on repeatability rather than absolute accuracy. You teach a point, and the robot goes back to it reliably. Problems start when that taught point drifts over time, when a new program needs to hit coordinates without manual teaching, or when a process demands tighter tolerances than the robot can deliver. Understanding which problem you actually have is the first step toward fixing it.

Common Causes of Robot Accuracy Degradation

Robot accuracy issues typically fall into a handful of categories. Knowing which one you are dealing with saves significant troubleshooting time.

Mechanical Wear and Backlash

Gearbox wear is the most common mechanical source of accuracy loss, particularly in robots that handle heavy payloads or run aggressive motion profiles. Harmonic drives and cycloidal reducers degrade over time, introducing backlash that manifests as position error—especially during direction reversals. If you notice that accuracy is worse in one axis than others, or that errors are direction-dependent, gearbox wear is a prime suspect.

Check for backlash by manually applying light torque to each joint with the servo drives disabled. Any perceptible play indicates wear. On robots with external measurement systems, you can also log following error during programmed reversals to quantify the issue. Proper servo motor sizing and selection at the design stage helps minimize long-term wear by keeping duty cycles within rated limits.

Calibration Drift

Every robot has a calibration reference—typically a set of encoder zero positions that correspond to a known mechanical configuration. Over time, these references can drift due to battery backup failures on absolute encoders, encoder hardware degradation, or physical impacts. A collision that moves an axis even slightly from its calibration mark will introduce a systematic offset on every subsequent move.

Calibration drift shows up as a consistent offset in one direction. If your robot is consistently 0.5 mm to the left, you likely have a calibration issue rather than a random mechanical problem. Most manufacturers provide a mastering procedure that uses reference marks, jigs, or laser measurement to re-establish the zero positions. This should be part of your preventive maintenance schedule.

Tool Center Point (TCP) Errors

The TCP defines the working point of the end effector relative to the robot flange. If the TCP definition does not match the physical tool, every calculated move will be wrong. TCP errors are common after tool changes, crash repairs, or when programming offline without physically verifying the tool dimensions.

A four-point TCP calibration method—where you jog the tool tip to a single reference point from four different orientations—remains one of the most reliable ways to verify TCP accuracy. If the calibrated TCP values differ significantly from your CAD model, measure the physical tool directly and investigate the source of the discrepancy. Bent tooling, worn gripper fingers, or incorrect mounting can all introduce TCP errors.

Thermal Effects

Robots generate heat during operation, and thermal expansion affects structural members, gearboxes, and even the base mounting. In high-duty-cycle applications, a robot that is accurate when cold may drift as it warms up. This is especially common in precision applications like laser cutting, dispensing, or small-part assembly.

If you suspect thermal drift, run a repeatability test at startup and again after two hours of continuous operation. Compare the results. Some manufacturers offer thermal compensation algorithms that model the expansion and adjust commanded positions in real time. In critical applications, allowing a warm-up cycle before production starts can improve first-part accuracy.

Compliance and Deflection

Every robot arm has finite stiffness. Under load, the structure deflects—and that deflection increases with reach. A robot holding a 20 kg tool at full extension will sag measurably compared to the same tool held close to the base. If your application involves varying loads or extreme reach positions, compliance may be your accuracy bottleneck.

This is a physics problem, not a maintenance problem. Selecting a robot with sufficient payload margin reduces deflection. Alternatively, designing the robotic cell so that critical operations happen closer to the base—where stiffness is highest—can mitigate the issue without upsizing the robot.

Systematic Troubleshooting Process

When accuracy problems appear, a structured approach prevents wasted effort.

Step 1: Quantify the Error

Before changing anything, measure the actual error. Use a dial indicator, laser tracker, or even a sharp pointed tool and a reference fixture to determine the magnitude, direction, and consistency of the deviation. Record whether the error is constant or variable, whether it depends on robot configuration, and whether it has changed recently.

Step 2: Isolate the Axis

If the error is predominantly in one direction, work backward through the kinematic chain to identify which joint contributes most. Locking individual axes and re-measuring can isolate the source. Software tools that display individual joint positions and following errors are valuable here.

Step 3: Check the Basics First

Before assuming mechanical failure, verify the simple things. Confirm that the mounting bolts are tight and the base is not shifting. Check that the tool is correctly mounted and not bent. Verify that the TCP and coordinate frames in the program match the physical setup. Review whether any program changes were made recently. A surprising number of accuracy complaints trace back to a modified frame or an incorrect tool number selection in the robot program.

Step 4: Run Standard Diagnostics

Most robot controllers include diagnostic routines—encoder checks, brake tests, motor current monitoring, and self-calibration procedures. Run these and compare results against the baseline values from commissioning. Elevated motor currents on a specific axis can indicate increased friction from wear. Encoder errors may point to hardware failures.

Step 5: Environmental Assessment

Check for external factors. Is the robot mounted on a surface that vibrates from nearby equipment? Has ambient temperature changed significantly? Is the electrical supply stable? These factors are easy to overlook but can have measurable effects on positioning performance.

Preventive Strategies

Preventing accuracy problems is more cost-effective than diagnosing them after they cause scrap.

  • Scheduled calibration verification: Check robot mastering and TCP calibration at defined intervals, not just when problems appear.
  • Collision detection: Enable the robot's built-in collision detection features. A detected collision should trigger a calibration check before resuming production.
  • Payload validation: Periodically verify that the actual payload matches the configured payload parameters. Changes in tooling or part weight that are not reflected in the robot configuration can affect both accuracy and servo performance.
  • Environmental monitoring: Track temperature at the robot base if your application is sensitive to thermal effects. Correlate any accuracy drift with temperature data.
  • Backup and documentation: Maintain records of calibration values, TCP measurements, and baseline accuracy data from commissioning. Without a known-good reference, diagnosing degradation is significantly harder.

When to Call for Support

Some accuracy problems are straightforward to resolve in-house—tightening a loose bolt, re-teaching a TCP, or correcting a frame assignment. Others require specialized equipment and expertise. Laser tracker-based calibration, gearbox replacement, or kinematic parameter identification are typically best handled by the robot manufacturer or an experienced integrator.

If you have quantified the error and isolated it to a specific cause but lack the tools or training to correct it, that data is extremely valuable to whoever handles the repair. The more information you can provide—magnitude, direction, axis, conditions—the faster the resolution.

Partner With AMD Machines

AMD Machines has over 30 years of experience building and maintaining robotic automation systems across demanding manufacturing environments. Our engineering team can help diagnose accuracy issues, perform calibration services, and design cells that maintain precision over the long term. Contact us to discuss your robot accuracy challenges.