Daimler tests robo taxis – on a platform from Bosch and Nvidia
One and a half years into their strategic cooperation to develop platforms for automated and autonomous driving (SAE Level 4/5), Daimler and Bosch have announced technical details of this platform. Nvidia is the supplier of the required AI platform – which comes as no surprise, as both Daimler and Bosch (as well as many other automakers and Tier One suppliers) use Nvidia’s AI computing platform for developments in connection with the automation of driving. Daimler has also recently introduced an AI-controlled infotainment system for its A-Class, which also uses Nvidia processors.
Artificial Intelligence (AI) is regarded throughout the industry as an important component for the complex control unit network of fully automated and driverless vehicles consisting of numerous individual ECUs. In addition to the Nvidia’s Drive Pegasus AI platform, the Nvidia scope of supply for Daimler’s self-driving cars also includes operating software. On the platform run the algorithms for vehicle movement generated by Bosch and Daimler using machine learning methods. In total, the ECU network achieves a computing capacity of several hundred trillion operations per second, Bosch said.
Automated driving requires a redundant system architecture and the highest level of functional safety, especially in urban environments. The ECU network therefore must meet these requirements. Here, the information from the various environmental sensors converges with radar, video, lidar and ultrasound technology. A video sensor, such as the stereo video camera from Bosch used here, generates 100 gigabytes of data per kilometer driven, for example.
The ECU network fuses the data from all environmental sensors, evaluates them within fractions of a second and plans the vehicle’s trajectory based on these data. This is done in real time with reaction times between 20 and 500 milliseconds. In order to achieve maximum functional safety, the necessary computing operations are carried out in parallel in different circuits. In the event of a fault, these parallel calculation results can be used without delay.
Due to the high computing capacity, these platforms require complex cooling. Towards this end, Bosch and Daimler have developed a sophisticated concept: Mercedes-Benz is planning battery-powered vehicles for the driving system to be jointly developed for fully automated and driverless driving in the city. The AI platform will therefore be integrated into the advanced cooling management system in the vehicle.
The tests are scheduled to begin in the second half of 2019. As a test region, the partners vaguely specified “a city in Silicon Valley”. According to Bosch, it has not yet been determined which city this will be. It is said that negotiations are still taking place. Both Bosch and Daimler conduct extensive development activities in Silicon Valley, each with its own company premises and shared development laboratories in Sunnyvale.
Daimler Mobility Services will operate the test fleet and the app-based mobility service. The pilot project aims to show how mobility services such as car sharing (car2go), ride hailing (mytaxi) and multimodal platforms (moovel) can be intelligently combined to shape the future of mobility.
Related articles:
Bosch, Daimler speed autonomous driving development
Daimler chooses Xilinx for AI development
Daimler tests automated driving in Bejing’s chaotic traffic
Bosch invests in Artificial Intelligence
How Bosch drives the development of the connected and the electric car
Daimler, Bosch demo driverless valet parking