Development and Evaluation of a Mobile Manipulation Robot for Surface Disinfection

Abstract:
Germs can infect people with potentially dangerous diseases, possibly leading to the death of the patients or an reduction of their quality of life. Thus it is important to regularly disinfect potentially contaminated surfaces. But due to a labour shortage in the cleaning sector, this task is sometimes not done frequent enough. Service robots equipped with disinfection tools can assist in this task and help to overcome the shortage. But existing disinfection robots come with restrictions due to the undirected application of chemicals or UV-C light and a missing cleaning effect of these methods. To tackle these issues, we developed the mobile robot “DeKonBot” that cleans and disinfects specific surfaces often touched by humans (door handles, door knobs, light switches/buttons). The robot is based on the Scitos X3 platform from MetraLabs GmbH, with an UR5e arm from Universal Robots A/S and an additional cleaning tool consisting of two brushes. To detect the surfaces for cleaning, we designed and trained a CNN model and projected these detections onto collected 3D data to precisely localize the surface relative to the robot. By adapting previously defined motion templates for each surface type to the extracted dimensions, we can then completely clean and disinfect the current surface. During an experiment in a partner’s office building, we evaluated the performance of "DeKonBot" by cleaning a total of 45 objects (17 door handles, 12 door knobs and 16 light switches). Overall only one error due to perception occurred during this experiment. On average, the disinfection process took 214s for door handles, 173s for door knobs and 164s for light switches. After additional optimizations, it was possible to further reduce the execution time of the motions by 50s.