NXP attacks Nvidia with number-crunching ADAS platform

In automotive electronics, two major trends determine future design approaches : First, the swelling data streams from environmental sensors demand a steep increase in number crunching capacity for the computers under the bonnet. Secondly, the variety of different control systems is increasingly reduced by combining the functions in domain controllers. With its BlueBox computing engine, NXP provides a hardware and software platform for such powerful domain controllers.

The BlueBox development platform, showcased at NXP’s Future Technology Forum (earlier known as Freescale Technology Forum), embraces a number of building blocks supporting the data infrastructure, number crunching and graphics power needed to turn highly sophisticated Advanced Driver Assistance Systems (ADAS) and eventually autonomously driving vehicles into reality. With all ICs that together form the BlueBox platform already available (just one is currently undergoing the requires qualification), the platform is almost series production ready, highlights NXP Chief Technology Officer Lars Reger.

With BlueBox a challenger to existing automotive high-performance computing platforms enters the stage. In particular its low energy consumption of just 40 watts eliminates the need for fans or liquid cooling, putting contenders like Nvidia into the shade. “ADAS need always to be able to detect the car’s environment at the lowest possible energy consumption”, Reger pointed out. “With energy budget of existing systems, we could easily run four BlueBox solutions, providing a redundant, fail-safe platform”, he added.

NXP's BlueBox: Sensor fusion, image processing and
safety controller on a single platform.
For full resolution click here.

At the heart of the BlueBox is a many-core chip developed for deployment in server farms. It contains NXPs S32V automotive vision processor well as the LS2088A embedded compute processor. The latter one in turn contains eight powerful 64-bit ARM Cortex-A72 number crunching cores along with specialized accelerators and high-performance communications interfaces that support and coordinate the massive data flow from multiple radar, camera, V2X and lidar sensors. The platform provides all the ADAS functionality necessary to enable automated driving, including interfaces to existing automotive data bus architectures like CAN, Ethernet and the like.

But BlueBox however is not only for autonomous driving and robot cars. “We likewise address the computing and sensor fusion needs of robotics in general”, Reger said. “In this industry there are many companies with terrific algorithms who are looking for the right platform for their projects”.

Related articles: