Auto-Assembly, Auto-mated

In vehicle manufacturing, entire subassemblies have long been fitted fully automatically. However, the systems used are often specialized to perform only a few or even just one single task. The resulting production lines are static and no longer meet the current requirements for assembly systems. Fact is, products are diversifying and markets are changing at an accelerating pace. This necessitates flexible systems that are highly mobile, adaptable and modular. This paradigm shift is not only a major challenge for automobile manufacturers, but also an exciting development area with a high potential for innovation. Together with Volkswagen Nutzfahrzeuge, researchers at Fraunhofer IPK have developed a model system for testing flexible approaches. Assembling interior vehicle paneling, a process that has been completely manual up to now, could soon be automated and serve as proof of the potential of autonomous assembly technologies.

Static solutions: a thing of the past 

One aspect of flexibility that was particularly important to the industrial partner is the solutions’ independence from location. The »Tend-O-Bot« robot system has already been used for a wide range of applications, from machine tending to human-robot collaboration. It consists of a mobile platform with an integrated robot arm and is able to independently navigate through the production facility thanks to modern mapping and localization algorithms. Additional sensors on board ensure that the mobile robot can react flexibly to obstacles and avoid collisions. Building on this technology, it can also reliably detect and localize assembly-relevant targets, in this case the vehicle body. Based on its visual assessment of the assembly situation, the system can even determine which assembly steps have already been carried out and which still need to be done.

After detecting the body and moving to the correct position, the Tend-O-Bot must precisely capture the assembly environment. In a multi-stage positioning process – from rough to fine – the system first estimates the location of its target based on AI models and refines it using conventional sensor processing at close range to achieve an accurate result. This enables the mobile manipulator to later approach the necessary assembly positions independently and flexibly with precision down to the millimeter.

The delicate touch of robotic hands

Now the actual assembly process can begin. The robot’s task is to clip a fine bracket on the back of the component into an assembly point – previously a manual process that required a high degree of tactile sensitivity. However, one of the specifications of the industrial partner is that the selected gripper technology must be able to handle a wide range of components with different geometries, even beyond this example task. The team of developers therefore relies on the use of a suction-based, flexible gripper, whose controls are integrated into the platform’s control system, to lift the component waiting to be mounted. The soft gripper adapts its shape to different geometries, thus enabling the assembly of a wide variety of components.

A crucial factor in the clipping process: Even the slightest spatial deviations inevitably lead to assembly failure and can potentially damage the vehicle body. This is why the process demands the highest standards of measurement accuracy and assembly process monitoring. Human workers can rely on their senses in situations like these, especially on haptic sensations and acoustic signals. Inspired by this tried-and-true human approach, the project team is evaluating the clipping process based on sensory feedback with an innovative procedure.

Complex interaction of modern technologies 

When many different hardware components come together, the potential for errors and interfacing difficulties is high. Wireless communication via 5G and the power supply on board the Tend-O-Bot pose additional challenges for the mobile solution. To meet the complex requirements of the task, the underlying software is both scalable and modular. Individual modules can be moved between the robot’s onboard computer and external cloud computers, which is ideal for coordinating signal runtimes, energy requirements and the utilization of available computing resources. Distributed development also leaves plenty of room for functionally expanding the solution in the future.

Closely integrating robotics, network technology, image processing and AI proved to be the key to success in the project with Volkswagen Nutzfahrzeuge. Using this real-world example, automotive manufacturers can now explore how even complex assembly tasks could soon become automated and autonomous.