Daimler and Bosch selected the solution of Nvidia drive for the parks of the robotaxi

On July 11, 2018 the company Nvidia announced that the companies Daimler and Bosch for the parks robotaxi were selected NVIDIA DRIVE. Daimler, Bosch and NVIDIA combined efforts to display on completely automated autonomous vehicles are expensive.

Daimler and Bosch selected the solution of Nvidia drive for the parks of the robotaxi

As noted in NVIDIA company, as a result of the announced cooperation of Daimler, Bosch and NVIDIA will create what in the autoindustry is called the 4th and 5th level of autonomy — cars which are capable to move on route independently. The allocated AI, cars will help to optimize traffic flows, to increase traffic safety and to make traficabilities even more well. The computing NVIDIA DRIVE platform will provide access to algorithms of deep learning, image understanding, cloud calculations and not only.

Cooperation of NVIDIA, Daimler and Bosch will integrate strengths of all three companies. NVIDIA will provide access to innovations in the field of AI and creation of platforms of autopiloting. Bosch, the supplier of automobile technologies, will introduce the examination in development of a hardware and system part. The Daimler company known to the world on Mercedes-Benz cars will provide the union with examination in the field of automotive industry in general and also will serve as guarantee of reliability and quality with which this world brand is associated.

In order that the trip to the robotaxi was safe and comfortable, the systems of the car should execute tens of difficult algorithms at the same time, literally for several milliseconds. For this purpose the enormous computing power is required.

Daimler and Bosch selected DRIVE Pegasus – the supercomputer with support of algorithms of AI developed especially for autonomous transport. The performance level in 320 TOPS (trillion operations per second) allows to cope with versatile and excess algorithms. Of the size of license plate number, the supercomputer provides computing opportunities the equal capacity of six workstations. It provides 1 trillion transactions on 1 Watt. Minimization of energy consumption allows to increase duration of operation of the equipment.

The Pegasus platform is designed to provide reliability and performance of unmanned vehicles. The supercomputer includes two systems - on - the chip (SoC) NVIDIA Xavier and two graphic processors developed for problems of AI and image processing. This hardware-software platform answers the level of automobile functional security according to the ASIL-D ISO 26262 standard. Even in case of detection of an error a system will be able to continue to function.

The solutions NVIDIA for the automobile industry include not only technologies which are used in the car. Supercomputers for AI of NVIDIA DGX are used in data centers for training of deep neural networks, allowing to allocate the car with capabilities of perception. The DGX-2 supercomputer with a performance of two petaFLOPS allows to train neuronets quicker and cheaper, than at servers of traditional architecture based on CPU, taking at the same time less places and consuming less electric power.

After training at powerful servers based on the GPU NVIDIA DRIVE Constellation AV simulator allows to test and certify software, intended for installation in the car. This program stack includes all aspects of piloting of autonomous transport: from detection of objects using algorithms of deep learning and computer vision before determination of location of the vehicle and making of a route. All this works at the DRIVE Pegasus platform.

"Partnership with Bosch and Daimler shows that the architecture of NVIDIA DRIVE Pegasus satisfies to necessary requirements of car makers for solution of the problem of autonomous driving. The combination of hardware solutions of NVIDIA for AI, the software, the integrated platforms and modeling tools and validation of results improves development process of autonomous transport". Luca De Ambroggi, the head of researches on AI in IHS Markit