Our guide on how to select an industrial robot briefly touched on the repeatability and precision (absolute accuracy) specifications of a robot. In this article, we present a deeper review around the topics of a robot’s positioning abilities and explain how inaccuracies can be mitigated by robot calibration.

The ISO 9283:1998 standard is a technical document which specifies the definitions and testing methods of a robot’s positioning performance, including accuracy, repeatability, deviation, reorientation, drift, overshoot, stabilization time, velocity characteristics, etc. Two of the most significant calibres are perhaps a robot’s repeatability and absolute accuracy. Repeatability is a robot’s ability to retain the same position or pose from time to time, and absolute accuracy means how well a robot can move to a specific position or pose in space.

The ISO 9283 Performance test specifies the testing procedure and conditions, which includes test path, environment, payload, number of cycles, etc. Most robot manufacturers only provide the repeatability specification of their robots, and do not reveal any information of how they conduct tests to draw conclusions. The absolute accuracy specification isn’t typically available from robot manufacturers.

Repeatability vs Absolute Accuracy

Industrial robots are known for their excellent repeatability, typically between 0.02 mm and 0.2 mm. A robot’s accuracy is generally 20 or more times worse than its repeatability, and this specification is unknown unless we conduct our own tests according to the ISO standard. Software tools like RoboDK can be used to automatically create ISO test paths, record accurate measurements from a high resolution measurement device e.g. a laser tracker, and generate reports of findings.

The complete repeatability and absolute accuracy specs include pose, distance, pose, path (see below).

1. Repeatability

• Unidirectional pose repeatability
• Distance repeatability
• Path repeatability

2. Accuracy

• Unidirectional pose accuracy
• Multi-directional pose accuracy
• Distance accuracy
• Path accuracy
• Path accuracy on reorientation

As a rule of thumb: the smaller a robotic arm is, the better its repeatability and absolute accuracy. Also, traditional industrial robots with their rigid builds generally perform better than collaborative robots of a similar size.

Source of Errors

But why industrial robots aren’t very accurate? There are a number of factors such as below.

• Intrinsic: Errors come from the components of a robot, such as part variations, non-linearity from dynamic influence of hysteresis and friction, resolution of joint position sensors, etc. Also overtime, the wear and tear of parts such as motors and gears.
• Environmental: Temperature and other external conditions.
• Computational: Controller errors such as small computer round-offs and steady state control errors.

Although errors come from various sources, they are not unexpected in modern machinery. Is it crucial to perform tests and attempt to remove them? The answer is it depends on the following aspects.

• Application – does the application rely on path accuracy e.g. straight lines have to be straight, or path repeatability e.g. it has to return to the same pick up point every time?
• Quality – is the robot made in high quality and still in a good condition? Also refer to the repeatability spec of the robot.
• Maintenance requirement – if a robot has gone through hardware replacement like a motor exchange, it is a good idea to calibrate it.

Robot Calibration

Finally, what are the calibration methods available to help improve a robot’s absolute accuracy?

Manufacturer Service: manufacturers like ABB and KUKA offer factory or on-site robot calibration as a service. This involves using a measurement device to take accurate measures per robot, and the result is reflected in the modification on the controller of the kinematics parameters of the robot. This changes how the controller transforms joint positions to end-effector positions in the Cartesian space, and vise versa.

Third Party Tool – Hardware: several commercial products are available which focus on ‘correcting’ the end-effector behaviour instead of the robot itself.

Third Party Tool – Software: different to the above two solutions, software approach focuses on not only the modification of a robot’s kinematic parameters, but also dynamic parameters such as stiffness to achieve a more complete kinematics model of a robot. RoboDK models 36 parameters for a robot’s kinematic calculations.