In the field of robotics, the requirements for positioning accuracy are highly demanding. Tolerances of components or workpieces increase the cycle time of robotic applications and make them error-prone. With the software ArtiMinds Learning & Analytics for Robots, processes with tolerances can be analyzed specifically and the robot program can be optimized based on the acquired sensor data. This significantly increases process reliability.

One of the biggest hurdles to robot-based automation are tolerances. Components itself often have high tolerances, but so do carriers, i.e. the carriers with which the products to be processed are moved into the cell.

Particularly in automation tasks with defined operating points, such as the insertion of wired components like large capacitors, coils, connectors or switches, this leads to a significantly higher cycle time when searching for the correct insertion position. In addition, the search becomes more error-prone, i.e. jamming or scratching of the surface can occur.

How can one generate a robot program that takes the high tolerances of the carrier into account - and does so without increasing the cycle time or, in the best case, even reducing the cycle time?
In the specific application, a robot was to pick up sensitive components at predefined fixed positions and place them on a product, which is transported into the cell with a carrier. The carriers provided for this purpose originate from batches produced in different plants and years. This resulted in high tolerances of more than +/- 1-2 mm. Due to the high financial investment already made for the carriers, it was decided to still use them for the task instead of purchasing new carriers with lower tolerances.
The robot was programmed using the software ArtiMinds Robot Programming Suite (RPS). Using a graphical interface, the user programs and configures his application via drag and drop of predefined function templates and can thus solve even more complex force-controlled tasks such as controlled joining or scanning of surfaces robustly and with little effort.

In order to reduce the cycle time of the force-controlled search, the working position, i.e. the position at which the search is started in the application, had to be optimized. The first step was to determine the optimum starting point. For this purpose, the application was analyzed with the additional software ArtiMinds Learning & Analytics for Robots (LAR). QR codes attached to the carriers allowed to scan them before entering the cell and to be uniquely identified. The search was initially run in unoptimized mode in order to determine the actual working position for each carrier. After only ten runs per carrier, the optimum starting position for the force-controlled search could already be read out from the LAR and that specifically for each carrier. Based on this data, the robot could simply be retaught and thus a time saving of up to 50% or 3 seconds per carrier could be realized compared to the unoptimized search.
The unambiguous identification of the carriers in the later running process and thus the reference for the robot, which starting point it should select for the search, takes place via the already integrated scanning process.

Additional TechInfo:
With ArtiMinds Learning & Analytics for Robots (LAR), a highly detailed and process specific insight into the application becomes possible. The basis is a wealth of data such as robot movements, force-torque measurements, image processing results and error codes, which are automatically collected and stored in a local database of the user. Instead of just providing general statements about the robot, ArtiMinds LAR is a tool making it possible to evaluate data in a targeted manner, with a direct view on the individual sub-processes programmed in the RPS. Robot tasks can be analyzed thoroughly and optimized with little effort.
A transverse force during gripping, for example, represents an error condition, whereas the same transverse force is desired during assembly. This contextual evaluation offers undreamt-of possibilities and simplifies the analysis of processes. An example would be the analysis of what effects different component batches have and how one can optimize teach points in a targeted manner in order to shorten cycle times and/or to increase process reliability.
As a web interface, LAR not only provides pure numbers and values, but also prepares them graphically and visualizes indicators. When analyzing the recorded data, meaningful evaluations are automatically suggested to the user facilitating interpretation.