Autonomous Driving: The Formula for Success

Have you ever wondered how much visual data humans are bombarded with every second? Researchers have calculated that our optic nerves transmit approximately one million bits of information to our brains every second. This calculation assumes that our retinas have already compressed this incoming information before it makes it to our brain’s gray matter.

Cars can also understand their environments in a number of ways. Advanced driver assistance systems (ADAS), which rely on cameras and radar systems, have made car travel significantly safer in recent years. Now manufacturers and suppliers such as ZF are working on the next stage: autonomous driving. When cars can drive themselves, human error – statistically the most likely cause of accidents – will no longer be an issue.

This requires an immense amount of computing power – one that standard control units currently do not possess. That is why ZF developed the ZF ProAI RoboThink , its AI-compatible automotive supercomputer that premiered at CES 2019, thus introducing the most powerful automotive control unit on the market to its product range. The latest model of the ZF ProAI product line is equipped with its own graphics processing unit, boasts a computing power of more than 150 tera-OPS (which is equivalent to 150 trillion operations per second), and can be modularly combined with up to four other units, resulting in a total performance of 600 tera-OPS. For comparison: in 1969, the Apollo 11 computer was able to process 41.6 operations per second.

The ZF ProAI platform’s scalability is particularly appealing to automotive manufacturers. “The modular hardware concept and the open software architecture are ZF ProAI’s biggest USPs,” says Torsten Gollewski, ZF’s Manager of Advanced Engineering and Design and Managing Director of Zukunft Ventures GmbH. The four models in the ZF ProAI product line can be optimally configured for any application – from basic ADAS functions all the way to self-driving cars and industrial applications. Customers can also choose their own software architecture. ZF is able to cover the entire scope of possible applications with four models:

Experts predict that ride-hailing services will generate the greatest demand for AI-compatible control units. Driverless robo-taxis are expected to reduce CO₂ emissions and gridlock in cities, thus saving time and helping the environment while still allowing for individual mobility. The demand for computing power in this field is particularly high. “Demand from ride-hailing service providers for even more computing power has arisen much sooner than predicted. The market for autonomous driving will be propelled forward by the new mobility providers much more so than by established vehicle manufacturers,” says Gollewski. The reason? In these mobility-as-a-service applications, vehicles not only need to process what is happening around them, which is already a complex task, but also need to integrate user data via the cloud, complete payment transactions, and, most importantly, determine and take the optimal route. Complex algorithms calculate these routes on the basis of the mobility and transportation requirements of passengers or goods, and compare them to the latest traffic conditions in real time. All of this is only possible with an extremely powerful mainframe computer – for example, the ZF ProAI RoboThink.