Advanced AI at the Edge
The robot now sorts the parts
Edge technology is rapidly evolving. This allows manufacturing companies to deploy even more sophisticated AI directly on-site more easily and quickly. An example of this is an innovative bin-picking system.
Predictive maintenance applications for maintenance or machine vision systems for quality assurance: Artificial intelligence is now widespread in factories and production halls. Advances in edge technology, such as more powerful IT platforms, open ecosystems, and specialized edge management platforms, now enable the next step. They allow manufacturing companies to equip their production facilities with more sophisticated AI, such as neural networks or generative AI, more quickly and easily.
Edge Ecosystem Brings Forth Innovative Bin-Picking Solution
An example of this is an innovative new bin-picking system that emerged from such an ecosystem. In this system, robotic arms pick various parts from a bin and place the identical parts at their designated locations. Unlike conventional solutions of this kind, the movements of the robotic arms are not pre-programmed but are carried out intelligently. AI software analyzes the images from a camera and controls the robot's grips. The progress is evident in the demo cell: Unlike statically programmed systems, the AI understands its environment and can take appropriate countermeasures, for example, if the parts are too close to the edge of the bin or to each other. In such cases, one of the arms shakes the bin to redistribute its contents.
Such a system can significantly help manufacturing companies and increase their efficiency. When parts of great variance need to be sorted and correctly assigned from a bin, this is usually done manually because suitable automation solutions have been lacking. The potential applications of the system range from machine loading to processing delivered parts for just-in-time production to intralogistics.
Separation of Control Hardware from Software
This solution is backed by an ecosystem of numerous partner companies. The control technology of the system uses ctrlX AUTOMATION from Bosch Rexroth and ctrlX OS. Unlike traditional industrial controllers, this is not tightly coupled with software in proprietary appliances but separates hardware and software. This allows the control system to be combined with various software depending on specific requirements. This approach enables it to function as a kind of "smartphone of automation." Companies can select and deploy the software they need from a comprehensive catalog. This catalog includes all the solutions from the partner companies in the ctrlX OS ecosystem.
One of the companies in this ecosystem is Vathos Robotics, whose machine vision software is used in the bin-picking system. This software analyzes the point clouds from a 3D sensor with a neural network and detects the type of object and its exact position in space. This enables the intelligent control of the robotic arms. Additionally, downstream systems can be connected to the software. For example, if the robotic arms have picked and sorted parts into a specific compartment, the system can automatically record this assignment in a connected warehouse management system.
A AI Server in the Control Cabinet
The IT platform of the system is a AI server from Dell Technologies, also a partner of Bosch Rexroth. This server runs the Linux-based real-time operating system ctrlX OS and the machine vision software from Vathos. The server is small and compact enough to fit into the system's control cabinet but still provides the necessary computing power to run the sophisticated AI software. As a "ruggedized" server, it is also specially hardened for use in industrial environments. It withstands extreme heat, dust, oil, shocks, and vibrations, conditions often found in factories and production halls.
MAIROTEC acted as the integrator of the overall system. The company specializes in automation, robotics, and plant modernization and is also a partner of Bosch Rexroth. The open ecosystem around ctrlX AUTOMATION allowed the integrator to combine the control system with Vathos' software without extensive development effort. This enabled the complex system to be implemented in a relatively short time and with little effort.
Quick Scaling and Comprehensive Security
Image source: Dell Technologies
The AI server used in the bin-picking system is a so-called "Native Edge Enabled Device." This means that the server can be used with the edge management platform "Dell Native Edge." This platform allows companies to centrally manage all their edge devices and deploy, manage, and secure the applications running on them across the enterprise. The platform integrates the solutions of smart factory equipment providers into its software catalog and offers blueprints for their automated rollout.
The comprehensive ecosystem of applications also includes the software solutions from Bosch Rexroth and Vathos. This gives companies the ability to quickly and easily scale the bin-picking system. If it is to be installed multiple times at several locations across the enterprise, the servers of all systems can be automatically equipped with the ctrlX OS operating system and Vathos' AI software from a central location.
Additionally, security mechanisms in the edge management platform ensure that the goal of Zero Trust can be significantly approached even in production. End devices for the platform are equipped with an operating system from Dell Technologies at the factory, and interfaces such as USB ports are deactivated. This prevents attackers from accessing the device to install infected software or create backdoors. Furthermore, the devices are provided with a digital voucher, without which commissioning is not possible. When a server is connected to the network, the edge platform performs a "security handshake" using the voucher. Only if it is the correct device and has not been altered does the platform deploy the software applications for the bin-picking system to the device.
Distribute Software for Training the AI to the Data Center or Cloud
Another advantage of the edge management platform is that it can also control the training of Vathos' AI software. To adapt the bin-picking system to their specific requirements, companies must first train the neural network with CAD models of the specific parts that the robots will pick and sort in the future. This training is too computationally intensive to be performed directly on the system at the edge. Dell Native Edge supports VMware and numerous Kubernetes distributions and offers integrations into the clouds of hyperscalers. This allows companies to distribute their computationally intensive tasks, such as training the artificial intelligence, from the platform to their data center or a cloud.
If you have any questions or require further information regarding ctrlX OS, please contact us:
