Making Vision-Guided Robotics Easy
Making Vision-Guided Robotics Easy
Until now, companies have rarely taught their robots to "see" due to fears of high integration costs and a compromise in process reliability. However, a robot control system based on Vathos' vision software, Dell Technologies' infrastructure solutions, ctrlX OS, and Bosch Rexroth's ctrlX AUTOMATION platform shows that it can be done differently.
For many years, "seeing" robots have been a relevant topic in automation. Especially in environments where the position of objects to be grasped is unknown (for example, when handling bulk material), image processing systems enable the automation of processes such as machine loading or assembly. Unlike fixed programming of object positions, the robot equipped with cameras reacts dynamically and flexibly to the random arrangement of the parts to be grasped.
Even though the advantage of image processing for robot guidance is quickly apparent, vision systems for robot control are still relatively rare in practice, given the multitude of automatable processes. This is primarily due to the high integration costs, as specialized knowledge is required for implementation. Additionally, a lack of process reliability deters many companies in practice, as robot control via image processing, despite its advantages, is often still considered unreliable.
With the advent of artificial intelligence (AI), which essentially involves the use of so-called neural networks, there is now an opportunity in robotics to overcome the two previous hurdles – namely, integration costs and process reliability. Through implicit learning from data instead of the tedious input of parameters into image processing programs, easy-to-use applications are now possible for the first time, without the customer or integrator needing specialized knowledge (keyword: "democratization of robotics"). The reduced error susceptibility and the ability to continuously learn changing environmental conditions, such as varying lighting conditions over time, simultaneously increase the chance of process reliability for such systems.
However, these newer systems are still not very common in practice. Primarily, these approaches are establishing themselves in quality assurance, but relatively rarely for the flexible control of robots, as the algorithm of the image processing program is an important but only a small part of the entire application or robot cell. More is needed to implement AI-based robot guidance in practice.
At Hannover Messe 2024, several partners came together at the Dell Technologies booth to demonstrate how intelligent robotics with AI can be implemented. The focus was on Vathos' modern vision software solution, which uses point cloud analysis from 3D sensors to precisely recognize objects learned by the AI and transmit the gripping poses to the robot. In the application case, three different components were classified, precisely grasped, and sorted into separate magazine rows. The software also checks each grip for potential collisions with other components or the edge of the bin based on the image data.
In AI, however, the underlying algorithm of the software is only one component. Equally important is that the software is built on the right foundation. The application requires a stable, industrial-grade operating system, must be able to learn many objects automatically – without human intervention – and, if necessary, share resources optimally over a network so that each robot does not need its own workstation with a screen and keyboard. Dell Technologies was convinced by the modern infrastructure and software architecture of the Vathos system, which opens up precisely these possibilities. Especially in enterprise environments, vision systems must be networked to continuously exchange process data and monitor machines – from anywhere, if necessary, even with a mobile phone. Dell Technologies provides the industrial-grade hardware and methods for such a system to securely distribute applications like Vathos' software to robot systems and have them managed by IT. Overall, AI in robotics is thus a very IT-driven topic.
However, the question also arises about the interaction between IT and the machine world. This is where ctrlX OS comes into play. The industrial operating system connects a multitude of applications with so-called Operational Technology (OT) and provides interfaces between robots and other components of a system, for example. MAIROTEC was found as an integrator that uses ctrlX OS and ctrlX AUTOMATION for its systems and can thus integrate Vathos' modern AI-based vision solution into its cells. At Hannover Messe 2024, MAIROTEC controlled two KUKA Agilus robots that received the workpiece poses from the Vathos system. The robot cell demonstrated how modern system integrators use the new AI methods for robotics and integrate them into a stable overall solution. Ultimately, the goal is to make these procedures easier to apply in practice, thereby effectively reducing integration costs and increasing the process reliability of vision-guided robots.
Image source: Dell Technologies
If you have any questions or require further information regarding ctrlX OS, please contact us:
