“The real world is far more complex than all scenarios with which autonomous vehicles have been trained so far. This is an exciting field of research with a variety of challenges! We have to disrupt existing training procedures in order to ensure flexible but also scalable mobility for autonomous driving. For this we need a revolutionary new approach to enable AI not only to learn but also to continue learning independently even without supervision. Because only a flexibly deciding artificial intelligence is truely intelligent! For this purpose, we would like to use the human being as a reference model. Because the abilities of humans must also be perfectly mastered by AI modules. A human being can always expand basic knowledge very flexibly and adapt to new situations by learning (delta learning). A car driver in Germany can adapt his driving experience to completely new traffic conditions after a short time, also abroad. This is exactly what we want to achieve with AI modules in autonomous vehicles. The AI of an autonomous vehicle must be flexibly expandable and must be in control of any situation. Only in this way we can optimally use the existing knowledge in changing situations. Regardless of whether the vehicle is driving in Germany or China, in winter or summer, in metropolises or rural areas.
Our KI Delta Learning project ensures both the expansion of the innovative power of the German automotive industry and a faster industrialisation of autonomous driving. This is how the German industry will become a global player in this future key technology. An exciting future lies ahead of us!“
Dr. Mohsen Sefati, Mercedes-Benz AG,
Project coordinator KI Delta Learning