Protected user data and technological basis

The visualization of the robot model in the 'Cobot Planner'. A simplified 3D model of the robot illustrates its state at the time of the collision and the position of the collision point.

© Fraunhofer IFF

User input may well contain confidential information. To protect it from unauthorized access, the Cobot planner does not save the input data. The server on which the web application runs is configured so that it irrevocably deletes all entries as soon as a user has been inactive for more than 30 minutes. To save the input data permanently for later reuse, the user can download the entire Cobot Planner workspace and save it locally. The next time the tool is used, it is then sufficient to upload the workspace again. This very simple data interface ensures that the data remains with the user.

The Cobot Planner's upload function also makes it possible to load models - for specific robots, for example - from other sources. This would eliminate the need to create a robot model yourself. The idea is for robot manufacturers to offer their own models for their robots, which users can load directly into the Cobot Planner. The Fraunhofer IFF is making an initial compilation of various robot models available free of charge at http://www.aroundrobots.com.

The plan is to continuously expand the Cobot Planner's range of functions until 2025. By the end of 2021, for example, it should be possible to switch between German and English. In addition, new body parts that are not yet selectable will be activated this year. The regular updates will also include selective improvements and corrections that users can report back to the Fraunhofer IFF via the Cobot Planner.

The technological basis

The technological basis for the Cobot Planner and its simulation section are various models that precisely simulate collision events and their effect on humans. The models include a hazard model, a dynamic robot model and a biomechanical model of the human being.

The hazard model is a structured data model that records the situation immediately preceding the collision on the basis of relevant characteristics and reproduces it for later simulation. The robot model takes into account the dynamic influence of the robot on the collision. It represents the robot as a multi-body system (MBS), which is parameterized before the simulation based on the user input. To make it easier for users to get started with the Cobot Planner, the Fraunhofer IFF has deliberately limited the input options to information that the user can also find in the robot's data sheet. A statistical regression model then estimates the missing model parameters from the inputs and transfers them to the robot model. The development of the regression model was preceded by a statistical analysis of the cobots available on the market. This showed that the numerous mass parameters of the FMD correlate with the total weight and maximum torques of the axes.

The author: Dr. Roland Behrens is Group Leader INS (Model-based HRC Integration and Safety) at the Fraunhofer IFF, Robot Systems Department, in Magdeburg.

© Fraunhofer IFF

Before each simulation, the Cobot planner reduces the complex MBS model to a sufficiently accurate substitute model. Thanks to the reduction, the otherwise very long simulation time is shortened while the accuracy of the results is maintained. Other parts of the model reflect the influence of the robot controller. As the control systems of the individual robot manufacturers differ greatly, the model is limited to basic control principles, such as those of safety functions. In principle, the Cobot Planner can also process robot models whose description contains all the necessary data and the exact control algorithms.

The human biomechanical model is based on data and findings from numerous and globally unique test subject studies conducted by the Fraunhofer IFF on behalf of the German Social Accident Insurance (DGUV) and the BGHM from 2015 to 2019. In the Cobot Planner, the human model calculates a separate force-deformation curve for each selected body part, as would occur in the event of a collision at the point of contact. Each characteristic curve then precisely reproduces the biomechanical behavior of the corresponding body part in the simulation. The cobot planner uses the robot model and the characteristic curves to simulate the collision separately for each selected body part. It then returns to the user the lowest speed that resulted for the most highly stressed body part.

The Fraunhofer IFF has experimentally validated the Cobot Planner's simulation results together with doctors from Otto von Guericke University's Department of Trauma Surgery. In the tests, test subjects were subjected to light impacts from a representative cobot. As soon as the test subject felt a slight sensation of pain above a certain collision speed, the ongoing test was terminated. The experiments took place with the involvement of the responsible ethics committee and under the supervision of doctors. The results show that the experimentally determined speeds scattered within the usual variance around the speeds determined by the cobot planner from the simulation.

Further development

The Cobot Planner technology is the result of the Fraunhofer IFF's preliminary work on the Computer-Aided Safety research focus. In this area, the Fraunhofer IFF has been working on practical and semi-automated methods for the reliable planning of HRC applications since 2017. The Fraunhofer IFF already has market-ready software solutions for power and force limitation as well as speed and distance monitoring (in accordance with DIN 10218-2 and ISO/TS 15066) that can be easily integrated into existing software for planning robot applications. The Fraunhofer IFF is currently working on further validating the technology used in the Cobot Planner. The medium-term goal of the work is to establish a reliable and fully digitalized alternative to collision measurement.