The next iconic wheel design could come from the mind of a computer... sort of.
Audi is now using artificial intelligence software that it has developed in-house to design new wheels. AI has long been touted as a solution to solving complex problems, and even the automotive industry can benefit. Porsche has used AI to detect noises, and Nissan is using the tech to build safer cars, but as the recent Instagram trend of unloading self-portraits to an AI-run app has shown, artificial intelligence can design things that are visually pleasing too.
Audi has long seen the value in advancing AI and even created a concept that uses the technology in its interactions with people, but the company is now using it more practically with a system it has dubbed FelGAN.
FelGAN is an in-house-developed AI-based software system that helps designers "discover motifs from completely new perspectives, giving them suggestions that they can further evolve and work into their creations." The software can "rapidly" propose a large number of photo-realistic designs on its own, or it can recombine existing designs "in a targeted way," meaning that the designers can tell the system what sorts of changes they'd like to see to a design and the software will provide various interpretations of those specific instructions.
Audi explains that the system effectively acts "as a kind of spontaneous idea hub for Audi's rim design team, allowing them to exchange new versions and variations." In this way, designers can easily see how changes in shape, color, surface structure, and other parameters would be realized in real-time.
The name "FelGAN" is derived from the German word for "rim" (Felge) and GAN, the latter of which is an acronym for Generative Adversarial Networks. GANs are "a special form of self-learning computer program in which two algorithms compete as opponents during the so-called training, becoming better and better in competition with each other."
One algorithm is the generator, creating artificial images of a specific motif, which in this case would be a vehicle's wheel. The other algorithm is the discriminator, or competitor, and this views a selection of images that is made up of real wheel photos alongside AI-generated images. The discriminator then decides which images show real wheels and which are generated. This process is repeated over and over until the training is over.
With this system, both algorithms are designed to learn from their mistakes and continuously improve, and after enough runs, Audi says that the generated designs are so real that the human eye cannot, or can only barely, tell the difference from those of real wheels. Everything is run on the cloud to avoid the use of high-performance hardware, and each design gets a mathematical value assigned to it, which the developers call "DNA." This DNA code can then be fed into the machine at any time to reproduce a design. In addition, Audi's human designers can feed the program with their own designs and photos. Audi says that the designers will often only use individual elements from a FelGAN design, and when satisfied, a prototype of the wheel is milled from plastic or aluminum. Basically, the human element is still very important.
In a world where even commercials can be written by artificial intelligence, we wouldn't be surprised if something like the e-tron GT is soon refreshed with AI-designed wheels.