In the past decade, artificial intelligence (AI) has gone through an amazing development. The promise that experts made, to make AI systems come close to human intelligence, however, has not been kept. This led to the so-called AI winter at the end of the 1980s. The fact that AI is experiencing a rapid renaissance is thanks, above all, to three supporting factors:
Hardware: Due to the gaming industry, graphic processors have become better and more affordable. At the same time, these types of processor are ideal for scientific calculations, which are performed on many levels at the same time.
Data: The Internet has made it possible to generate and distribute large annotated* and publicly available data sets. One of the first was so-called ImageNet. Hundreds of additional data sets followed.
Software: The algorithms for training very large neural networks have been developed further and, today, are summarized under the term “deep learning.” Thanks to deep learning, the results of the existing, conventional algorithms are significantly exceeded in many areas, such as in image recognition, speech recognition, translation and understanding text.
In aviation, deep learning has been on the back burner to date, although the potential it holds when it comes to the optimization of product and maintenance processes is huge. With the set-up of the AI Aviation Lab, we are creating a catalyzer at ZAL for the development of deep learning prototypes for the aviation industry. At the same time, space is given to focus on the potential of deep learning in aviation. The Lab is supported by the AI team of experts at ZAL GmbH, which combines domain knowledge in aviation with AI know-how.
The AI Aviation Lab offers your company equipment and services along the chain create – sense – compute – act – explore.
For generating ideas and holding workshops, a creative corner is featured, consisting of a large whiteboard and seating.
To record training data, various sensors are available, such as a high-speed camera, 3D cameras, Lidar and a high-end microphone. Via a high-performance workstation laptop, the data can be captured.
For training neural networks, the AI Aviation Lab is equipped with a GPU cluster, which contains four of the highest-performing graphic processors (NVIDIA Tesla V100) to date. The cluster is situated in a cooled rack to protect it from overheating.
A modern robotic arm from the brand Universal Robots (UR10e) can be integrated into each case of application. This is additionally outfitted with a three-finger gripper and a wrist camera from Robotiq. On top of this, the AI Aviation Lab contains a ROSbot 2.0 from Husarion for testing and teaching purposes.
The lab features two visualization options for examining project results and for presenting demonstrations. An 86-inch touch monitor is available that enables the user to access all the devices in the lab – while a 6 x 3-meter sheet of glass is available for presenting data in 4K through rear projection and for enabling novel presentation methods.
Thanks to these five elements, the AI Aviation Lab makes it easier for experts to quickly develop proof-of concepts, and it enables newcomers, too, to get started on the topic.
Would you like to use the AI Aviation Lab to gain initial experience in the application of industrial artificial intelligence for your company? Please feel free to contact us about bookings and for additional information at:
*Annotated data sets comprise machine-readable information about the content of each data set and are used to train algorithms for artificial intelligence. For instance, each identification of an object can contain information about the type or location of the object in the room.