What Is the Difference Between Robot Accuracy and Repeatability?
Robot accuracy measures how closely the robot reaches a commanded position compared to the true target location, while repeatability measures how consistently the robot returns to the same position over multiple attempts. Accuracy indicates absolute positioning error relative to a reference coordinate system, whereas repeatability indicates precision when returning to a taught point.
Understanding this distinction is critical for robot applications. A robot might have poor accuracy (commanded positions are offset from true locations) but excellent repeatability (it returns to the same offset position consistently). For most industrial applications, repeatability matters more than accuracy because processes are taught by moving the robot to desired positions rather than programming absolute coordinates.
Typical industrial robots achieve repeatability of ±0.02 to ±0.1 mm while accuracy might be several millimeters without calibration. This explains why teach programming works well despite accuracy limitations. When you teach a robot to pick parts from a fixture, you're recording the actual position the robot achieved, not commanding theoretical coordinates. As long as the robot returns to that same position reliably, the application succeeds.
What Is the Difference Between Robot Accuracy and Repeatability?
Accuracy and repeatability represent fundamentally different performance characteristics that determine robot suitability for specific applications.
Accuracy vs Repeatability: Feature Comparison
| Feature | Accuracy | Repeatability |
| Definition | How close the robot gets to the commanded target position | How consistently the robot returns to the same position |
| Measurement | Deviation between commanded and actual position in 3D space | Variation in position across multiple attempts to reach the same point |
| Typical Values | ±1 to ±5 mm for industrial robots (uncalibrated) | ±0.02 to ±0.1 mm for industrial robots |
| Affects | Absolute positioning, coordinate accuracy, multi-robot coordination | Process consistency, part-to-part variation, taught position reliability |
| Can Be Improved By | Kinematic calibration, external measurement systems | Better mechanical design, encoder resolution, temperature control |
| Critical For | Offline programming, model-based path planning, vision-guided positioning | Traditional teach programming, repetitive operations, precision assembly |
| Changes Over Time | Worsens with wear, temperature, and mechanical changes | Relatively stable until significant wear occurs |
| Cost to Optimize | Expensive (calibration equipment and time) | Built into robot design, difficult to improve after purchase |
Visualizing the Difference
Consider a target at coordinates X=500mm, Y=300mm, Z=200mm. You command the robot to move there 10 times:
- High Repeatability, Low Accuracy: All 10 attempts land within ±0.05mm of each other, but they're clustered at X=503mm, Y=298mm, Z=202mm. The robot is 3-4mm off target but extremely consistent.
- High Accuracy, Low Repeatability: The 10 attempts average out to X=500.5mm, Y=300.2mm, Z=200.1mm (very close to target), but individual attempts scatter across a ±2mm range. The robot gets close on average but isn't reliable.
- High Accuracy and Repeatability: All 10 attempts cluster within ±0.05mm of X=500mm, Y=300mm, Z=200mm. This represents ideal performance but is rare without calibration.
Why the Difference Exists
Repeatability depends primarily on mechanical precision: gear backlash, bearing tolerances, encoder resolution, and structural rigidity. These factors determine how consistently the robot achieves the same joint angles, which translates to consistent end-effector position.
Accuracy depends on how well the robot's control system understands the relationship between joint angles and Cartesian space. This requires precise knowledge of link lengths, joint offsets, gear ratios, and mechanical compliance. Manufacturing tolerances mean each robot has slight variations from the nominal kinematic model, causing accuracy errors even when repeatability is excellent.
Why Is Repeatability More Important Than Accuracy?
Repeatability matters more for most applications because industrial robots are programmed by teaching actual positions rather than entering theoretical coordinates, making consistent return to taught positions critical while absolute accuracy becomes secondary.
Teach Programming Methodology
The vast majority of industrial robot programming uses teaching methods where operators manually guide the robot to desired positions and record those locations. When teaching a pick point, you move the robot until the gripper aligns with the part, then save that position. The control system records the joint angles achieved, not the theoretical Cartesian coordinates.
When the program runs, the robot reproduces those same joint angles, returning to the physical position you taught. As long as repeatability is high, the robot returns accurately to the taught location even if that location doesn't match the theoretical coordinates calculated from the kinematic model. Accuracy errors don't matter because you never relied on the theoretical model.
This explains why robots with ±3mm accuracy but ±0.05mm repeatability successfully perform precision assembly. The absolute position might be offset from the commanded coordinates, but the robot consistently returns to the exact same spot every cycle.
When Accuracy Becomes Important
Offline Programming: When programming robots in simulation software without physical teaching, accuracy matters. The simulated robot must behave like the real robot for programs to transfer correctly. Poor accuracy means significant program adjustments after download.
Vision-Guided Applications: When a vision system calculates part coordinates and commands the robot to move there, accuracy determines whether the robot reaches the correct location. Vision systems provide absolute coordinates in Cartesian space, not taught joint positions.
Multi-Robot Coordination: When multiple robots work on the same part or share workspace, their coordinate systems must align accurately. Poor accuracy on either robot causes misalignment even if both have excellent repeatability.
Model-Based Path Planning: Applications using CAD models to generate robot paths require accurate transformation between the CAD coordinate system and the robot's working frame. Accuracy errors cause programmed paths to miss intended features.
Accuracy Improvement Through Calibration
Robot accuracy can be improved from ±3mm to ±0.3mm or better through kinematic calibration. External measurement systems (laser trackers, coordinate measuring machines) measure the robot's actual positions across its workspace. Software analyzes these measurements to determine the robot's true kinematic parameters, updating the control model.
How Do Robot Manufacturers Measure Repeatability?
Robot manufacturers measure repeatability using ISO 9283 standards, which specify testing a single position 30 times and calculating the radius of a sphere containing all measured points, with smaller radii indicating better repeatability.
ISO 9283 Test Procedure
The standard requires testing five positions distributed throughout the robot's workspace. For each position, the robot moves away and returns 30 times. A precision measurement device (typically a laser tracker or coordinate measuring machine) records the actual end-effector position for each return.
The 30 positions form a cluster in 3D space. Repeatability is defined as the radius of the smallest sphere centered on the average position that contains all 30 points. This metric, called pose repeatability (RP), represents the maximum deviation from the average position.
Testing occurs under controlled conditions: consistent temperature, no payload changes, and the robot at normal operating speed. Manufacturers typically report the worst-case repeatability observed across all five test positions.
Unidirectional vs Multidirectional
ISO 9283 specifies unidirectional testing where the robot approaches each test position from the same direction every time. This represents best-case performance and the values manufacturers typically advertise.
Multidirectional testing has the robot approach from different directions, revealing backlash and mechanical play. Multidirectional repeatability is worse than unidirectional, sometimes by a factor of 2-3x. Real applications often involve approaches from varying directions, so multidirectional performance better predicts actual capability.
Real-World Considerations
Manufacturer specifications represent new robots under ideal conditions. Repeatability degrades over time due to gear wear, bearing deterioration, and mechanical loosening. A robot specified at ±0.05mm repeatability might drift to ±0.1mm after several years of intensive use.
Temperature variations affect repeatability. A 10°C temperature change can cause expansion or contraction that impacts position by 0.1-0.2mm on a large robot. Applications requiring extreme repeatability need temperature-controlled environments.
Payload and speed also influence repeatability. Testing occurs at rated payload and standard speed, but exceeding these specifications or running at maximum speed degrades performance. Dynamic effects during high-speed motion introduce position errors that static testing doesn't capture.
What Factors Affect Accuracy Over Time?
Accuracy degrades over time due to mechanical wear, temperature variations, structural compliance under load, and changes in kinematic parameters from component replacement or collision damage.
Mechanical Wear
Gear backlash increases as gears wear, creating slack in the transmission between motors and joints. This introduces position errors that worsen gradually over thousands of operating hours. Harmonic drives, common in robot joints, experience wear that causes lost motion and position drift.
Bearing wear allows increased play in joints, particularly in larger robots with higher moment loads. The wrist joints, experiencing the most dynamic loading, typically show wear effects first. Regular inspection and timely bearing replacement maintain accuracy.
Temperature Effects
Robot structures expand and contract with temperature changes. A 2-meter reach robot arm might change length by 0.2mm per 10°C temperature variation due to thermal expansion of aluminum or steel structures. This directly affects accuracy since the control system assumes constant link lengths.
Motors and gearboxes generate heat during operation, causing thermal gradients across the robot structure. A cold robot moved to a working position will be at a slightly different location once thermal equilibrium is reached. Temperature-sensitive applications require warmup periods or active temperature compensation.
Structural Compliance
All robot structures flex under load. Extending a heavy payload to maximum reach causes measurable deflection at the tool center point. The control system compensates for gravity-induced deflection based on the configured payload, but payload errors or dynamic effects during motion introduce accuracy errors.
Repeated high-stress operations can cause permanent structural deformation. Robots regularly operating at maximum payload or experiencing frequent abrupt stops may develop accuracy drift as structural components take permanent set.
Calibration Drift
Even calibrated robots experience accuracy drift over time. The kinematic parameters determined during calibration gradually become less accurate as wear and temperature effects accumulate. High-precision applications requiring sustained accuracy need periodic recalibration, typically annually or after major component replacement.
Collisions immediately invalidate calibration. Any impact that moves or bends structural components changes the kinematic parameters. Post-collision recalibration is essential for applications depending on accuracy.
How Do Accuracy and Repeatability Compare Across Robot Types?
Six-axis articulated robots typically achieve ±0.03-0.08mm repeatability with ±2-3mm accuracy, SCARA robots reach ±0.01-0.02mm repeatability with ±0.1-0.5mm accuracy, and delta robots deliver ±0.05-0.1mm repeatability with ±0.3-1mm accuracy.
Six-Axis Articulated Robots
These robots accumulate position errors through six serial joints, making accuracy challenging. Each joint's positioning error compounds with errors in subsequent joints. Typical uncalibrated accuracy is ±2-5mm, though repeatability reaches ±0.03-0.1mm depending on size and quality.
Larger robots have worse accuracy due to longer kinematic chains and structural compliance. A small 500mm reach robot might achieve ±1mm accuracy, while a 3-meter reach robot could have ±5mm accuracy before calibration. Repeatability scales less dramatically, with large robots still achieving ±0.08-0.1mm.
SCARA Robots
SCARA robots excel at accuracy and repeatability in their horizontal working plane. The rigid vertical axis and shorter kinematic chain (four axes vs six) reduce error accumulation. Typical accuracy is ±0.1-0.5mm with repeatability of ±0.01-0.02mm.
The SCARA's natural compliance in horizontal directions combined with rigidity in the vertical axis creates predictable behavior that simplifies calibration. Many SCARA applications achieve excellent results without kinematic calibration.
Delta Robots
Delta robots use parallel kinematics where three arms share positioning load. This parallel structure provides better accuracy than serial kinematic robots but still exhibits position errors from manufacturing tolerances and structural compliance. Typical accuracy is ±0.3-1mm with repeatability of ±0.05-0.1mm.
The lightweight moving structure contributes to good repeatability, but the complex kinematics make accuracy calibration more challenging than SCARA or six-axis robots. Dynamic effects during high-speed operation can introduce position errors beyond static specifications.
Collaborative Robots
Cobots typically have slightly worse accuracy and repeatability than comparable industrial robots due to their force-limiting joints and safety-focused designs. Typical repeatability is ±0.05-0.1mm with accuracy of ±3-5mm. The compliant joints necessary for collision detection introduce some position uncertainty.
However, cobots' excellent repeatability still enables most precision assembly and handling tasks. The slight performance trade-off is acceptable given their safety features and ease of deployment.
Conclusion
Understanding the distinction between accuracy and repeatability is essential for successful robot applications. Repeatability determines how consistently a robot performs taught operations, making it the critical specification for traditional industrial applications using teach programming. Accuracy matters for offline programming, vision-guided operations, and multi-robot coordination where absolute position correctness is necessary.
Most industrial robots prioritize repeatability over accuracy, achieving impressive consistency (±0.02-0.1mm) while maintaining moderate accuracy (±1-5mm). This performance profile matches typical application requirements where teaching positions compensates for accuracy limitations. Applications requiring both high accuracy and repeatability can achieve them through kinematic calibration, though at additional cost and complexity.
Recommended Resources
Explore more robotics insights and industry developments
ABB launches groundbreaking Ultra Accuracy for GoFa cobots
ABB Robotics today launched Ultra Accuracy, a new industry-leading feature for its GoFa cobot family that delivers the highest level of precision available in cobots, enabling over 10 times greater path accuracy vs. other cobots...
Reducing the Variables: Getting Accurate Motion From Inaccurate Robots
In automotive, repeatability is everything. The industry has evolved its production practices over decades to play to the strengths of the robots they use, and avoid the weaknesses.
How to Deploy Efficient Pick And Place Robotic Arm
A precise pick and place robotic arm enables an efficient performance due to accuracy and speed of operations. Collaborative robots or cobots ensure fast, stable, and safe deployment of manufacturing tasks.