The goal in intralogistics is autonomous vehicles that transport goods from point A to point B independently. As a rule, automated guided vehicles (AGVs) use several sensors that work with different principles: RGB cameras, 3D cameras, laser scanners and radar or ultrasonic sensors. They represent the sensory organs of the mobile machine, so to speak, whose information is compiled through sensor data fusion, i.e. the combination of data from the various sensors. However, the synchronization of the sensors and the fusion of the sensor data pose major challenges for users and hardware.

ifm has developed the O3R hardware platform specifically for this area, which makes both preliminary and series development easier for developers thanks to a smart software environment and an extensive selection of software tools and interfaces.

The central component is an edge device that provides both high computing power and the ability to easily connect a wide variety of sensors. A total of up to six 3D cameras and numerous other sensors can be connected. The camera connection is made via FDP-Link (Flat Panel Display Link), while GigE interfaces are available for the other sensors. CAN interfaces ensure simple integration into the architecture of a mobile robot. A Linux system equipped with an NVIDIA video processing unit forms the hardware basis. As the performance of this GPU is scalable, it can be adapted to the respective application. With the available ROS2 drivers, the system can be integrated into robotics applications. And since the image processing in the O3R concept moves to the edge device, hardly any data processing is required in the camera itself. This means that several different cameras can be used. ifm offers suitable camera heads that contain 3D sensors or a combination of 3D and 2D sensors with different aperture angles and resolutions.

As the Linux-based edge device offers a great deal of computing power, even sophisticated applications can be implemented according to the provider; artificial intelligence (AI) applications in particular rely on this performance. For example, neural networks can be implemented for image processing applications that would not be possible using conventional algorithmic methods. These deep learning methods can be used, for example, to improve the orientation of autonomous mobile robots (AMR). The method used for this purpose, simultaneous localization and mapping (slamming), enables the AMR to 'know' what its environment looks like and where it is located within this environment (localization). If it moves within this environment, it can also create a map of its surroundings (mapping). Such tasks can be solved by using neural networks and other AI methods.