In southern Germany, in the state of Baden Wurtemberg, Daimler and Bosch have built a fortress, laboratory and research center where they are cooking what the future autonomous car will look like. After an hour and a half drive from Zurich, and less than 200 kilometers from Sttutgart, is this enclave that Daimler began to build in 2015 and that has a team of about 300 engineers from both companies and specially designed road areas to test and test the developments that are carried out there and that are integrated into the vehicles of the Teutonic brand.
In this idyllic environment, surrounded by greenery, the combination of images provided by the latest generation RADAR and LIDAR sensors created by Bosch are put to the test, as well as those captured from high-resolution and infrared cameras that equip Mercedes Benz cars from Daimler. All those images, and the data they provide, are merged and mixed to be processed and computed under NVIDIA’s Pegasus platform, which applies artificial intelligence (AI) and deep learning technologies and neural networks to decipher in a matter of milliseconds what it is. what the images say and what decisions must also be made in real time based on them.
A revealing data shows the relevance of the tests that Daimler and Bosch are developing in Inmmendingen, and that will continue to be carried out until reaching the second phase of the project in the second half of 2019 in the San Francisco Bay. The autonomous cars that are tested in the German town are integrated with eight RADAR sensors, which have the peculiarity of using radio waves to detect and send the data it identifies to the AI platform from about 250 meters away. Only those eight sensors provide 100 Megabytes of information every second, which would be equivalent to almost 1 Gigabyte of information if all the RADAR sensors are added, every second. Or, what is the same, about 60 Gigabytes per minute, 360 per hour, or 8 Terabytes of information every day.
And these figures are only those provided by RADAR sensors. Daimler and Bosch do not offer details of how many LIDARs, which use lasers to capture in 3D images everything around them in 360 degrees. Nor the amount of data obtained from the high-resolution cameras and ultrasound that equip the vehicles. In any case, all this is computed and analyzed in the neural networks and deep learning capabilities that the NVIDIA platform develops.
Through it, the variety of incoming data is analyzed and interpreted, and translated into driving commands in a very short time thanks to the computational power of the NVIDIA AI platform. Daimler and Bosch also report in Inmmendingen how fully automated and chauffeured driving in the city requires a redundant and versatile system architecture and the highest level of functional safety. To achieve this level of security, the necessary computing operations are performed in parallel on different circuits. And this means that the system has immediate recourse to these parallel comparative results when necessary.
The remaining months between now and the second half of 2019 to reach Metropolis in California, the pilot city where Daimler and Bosch’s fleet of automatic tests will roll, seem exciting. If all the sensors that manage the cars that Daimler tests in Inmmendingen are kept at the same level of data computation up to date, and if I have done the math correctly, within a year the equivalent of 3 Zettabytes of data would have been processed, that would be the same capacity that 150 million iPhone Xs of 256 GBytes each would have today. And just taking the eight RADAR sensors as a sample.
Daimler and Bosch will also offer automated transportation service on select routes in the second half of next year in a city located on Silicon Valley’s San Francisco Bay. This test operation will provide information on how fully automated and chauffeured vehicles can be integrated into a multimodal transport network. Many cities face numerous challenges that are increasingly weighing down the existing transportation system. The test is to show how this new technology could be a solution to these challenges. Will we see it? At Inmmendingen we have seen it, and we have tested Mercedes Benz driverless cars. A year from now, in San Francisco, …
