Is it possible to polish to a mirror finish using a cobot?

Good polishing experts are hard to find and that is not about to change. The work requires knowledge and experience, is time-consuming, expensive, monotonous and stressful, and takes place in an unhealthy environment. The need for experts is especially acute in the mould construction business.

Sirris partnered with the German Fraunhofer Institute for Production Technology (IPT) to find out whether semi-automation using cobots could take (some of) the burden off polishing experts. The Fraunhofer IPT is renowned for its research into the automation of polishing processes, mainly investigating automation using industrial robots and actively controlled adaptive tool holders.

Sirris set up a cobot polishing cell to research whether a cobot can be usefully deployed as a polishing tool for experts to use.


Semi-automation of polishing using cobots is a new approach that will not make polishing experts redundant but – if it can be deployed successfully – could relieve them of the tedious, repetitive sanding of large surfaces. This would free them up to focus their expertise and dexterity on finishing the intricate elements of the mould parts, such as small radii, hard-to-reach places and complex shapes.

Questions explored in the research include:

  • Can the force measurement inherent in a cobot also be used to polish a surface to a mirror finish without the cobot's control cycle leaving traces in the surface?
  • To what degree of shape complexity can the cobot be deployed in a user-friendly way? (To determine this, complexity is gradually increased from flat to large radius and then small radius.)
  • What approach could a polishing expert use to teach the cobot new tasks in a user-friendly manner: teach-by-demonstration, external sensors, offline programming, etc.? 

Test results

The specific tests have so far focused on flat and slightly curved surfaces. In the final phase of the research, more significantly curved surfaces will also be examined. Teach-by-demonstration programs have been created so that the operator can 'show' the cobot where to polish the surface. Moreover, configurable parameters ensure the specific magnitudes for the polishing process, such as feed, speed, contact pressure, angle of attack or overlap, can be quickly adapted based on operator assessment.

As a linear movement with a rotating tool almost always leaves visual traces in the reflective surface, tests were carried out to determine which superposed movement, if any, best masks these visual traces. To this end, the results of five strategies were compared, both visually and based on measurements under a KEYENCE VK-X1100 confocal microscope. The images below show the results of the tests.


The preliminary test results show that the internal force-feedback control loop of the KUKA iiwa cobot can be suitably deployed to polish to a mirror finish.

Through teach-by-demonstration, the operator can operate the cobot without having to get to grips with the complexity of robot programming. Further steps are needed to arrive at a finished program that allows sufficient flexibility and is industrially deployable.

A subsequent test will also explore the possibility of autonomously machining 2D curved surfaces in a similar way. Since the cobot follows the surface in compliant mode, in which a constant pre-set contact pressure is maintained, the previously performed polishing trajectory can be used each time as a newly taught trajectory and the normal vector on the 2D curved surface can be calculated at each point. For this, the cobot will only need to be taught one start trajectory, together with the edge contours of the surface to be polished, which will also make learning a new 2D curved surface for polishing very user-friendly.

The research shows that cobotised polishing has potential. Over the next few years, Sirris will continue to work on improving these processes and further exploring the possibilities of cobotised finishing (e.g. sanding, deburring).

Partner Name:
Sirris Logo