Select your location:

Location

Select your language:

Opera

By developing probabilistic models, the OPERA grant project is working to describe the accuracy of the robot and its environment.

Collaborative robots, so-called CoBots, have enabled manufacturing processes to be set up intuitively through hand-guided programming. Medium-sized companies in particular benefit from this development by adapting an exploratory approach. However, this procedure involves a number of finesses in terms of accuracy and error tolerances, so that until now, despite simple handling, an expert usually had to be consulted during programming. had to be consulted for the programming.

The OPERA project is working to change this situation by providing the end user with intuitive tools to analyze and optimize the reliability of manufacturing processes and work cells. To this end, probabilistic models describing the accuracy of the robot and its environment are being developed and exemplarily integrated into the latest generation of KUKA robot controllers. Together with DLR as well as our associated partners Zollner and Rational the developments are carried out and evaluated on site using relevant use cases.

Modeling the robot and environment inaccuracies – a paradigm shift

In industrial robotics, the focus is on building and using robots as error-free positioning devices to be able to ignore inaccuracies. In the case of CoBots in particular, however, one must be aware of inaccuracies in the long term and learn how to deal with them. However, since not all sources of error can always be determined deterministically, probabilistic models are suitable here.

Ideally, the position of the robot arm and the gripper as well as the workpiece would be always known exactly in an application. In reality, the gripper on the robot as well as the object only have a certain probability of being in the vicinity of the assumed position, and the actual position is not known exactly. So far, the robot expert has the data sheets of the robot and the gripper in his head and can thus estimate from his experience how precisely the tasks can be solved.

In the OPERA project, we are now working on this one step further and not leaving the user alone with the partially incorrect ideal image, but making the uncertainties visible and assessable in a 3D model. In addition, the OPERA project will make it possible for the first time to take inaccuracies into account in the form of probabilistic models in process sequences. This should enable increased flexibility, accuracy and reliability in complex tasks.

Opera goals

The goal is to transparently represent model parameters and model errors in the kinematics and dynamics of CoBots as well as the robot environment and make them available to the user so that system integrators and end users can more intuitively and better estimate the performance of the robot system. With the representation of the error potential of the robot system, it should furthermore be possible to make predictions regarding the influence of model accuracies on the feasibility of individual work steps, so that semantic knowledge can be taken into account in the process design in addition to geometric variables. Ultimately, this should make it possible to automatically design manufacturing processes and the associated work cells for increased reliability by using model parameters and model errors, but also their influence on individual actions of a sequence, for optimization.