This is the brain of the autonomous car: a supercomputer that fits in a shoe box

Still more coming news CES 2016, in which the clear protagonist is the autonomous car. To get a car to be able to drive by their own means, without a driver, it is necessary to process a huge amount of information. Information collected by sensors, positioning systems, and an accurate recognition of images that is able to differentiate between a traffic signal, a pedestrian, or a car that just skipped a traffic light. That is the reason why the autonomous car could not be understood without a computer system really powerful and advanced, that also has to be robust and reliable, so critical that it will be the task that has been assigned. And it is there where he has entered in the game NVIDIA, more known for its graphics cards, to launch the system DRIVE PX 2, a supercomputer that fits in a shoe box.

NVIDIA specialized systems DRIVE PX 2, intended for the automobile industry, in paradigms of machine learning and neural networks, in solutions of image recognition and Artificial Intelligence aimed at cars connected and autonomous.

beyond the power that require these systems, NVIDIA has been working for a long time in GPUs, graphics processing units, based on paradigms of machine learning, and neural networks. Christian, and without going into too much detail, this technology would allow to create systems of recognition of images, for example, to detect pedestrians, each time more precise and efficient, able to detect a pedestrian with a smaller number of calculations. A technology that has improved significantly in recent years, thanks to initiatives such as the ZMP in Japan, which is developing a kind of taxis, self-employed, without a driver.

The system NVIDIA DRIVE PX 2 will begin to be incorporated in the models of many manufacturers who are already customers of the technology of Santa Clara, California, and among those who are already Audi, BMW, Ford, Mercedes-Benz, or Volvo, which has already confirmed that this system will be the responsible of the independent abilities of the 100 Volvo XC90 “without driver” which will begin shooting next year in Sweden.

NVIDIA-drive-px2-01


  • This compact system that fits in a shoe box, it is able to run 24 trillion of operations of deep learning, such as 150 MacBook Pro.

    NVIDIA DRIVE PX 2 is able to run 24 trillion operations algorithms of deep learning per second, the equivalent of 150 laptops MacBook Pro. Among other things, because its architecture has been optimized to run this type of algorithms and accelerate the development of Artificial Intelligence.

    not to mention another really important aspect, which not only affect the self-employed, but the cars are connected, in general. Beyond the amount of data that you collect a car from sensors and cameras, the future is to create millions of mind-artificial, and with four wheels, connected in the cloud. That is to say, that all such data, and that all of that “experience” accumulated by each car, to be shared with other vehicles, in pursuit of a safer driving and to get the autonomous car to be feasible and reliable.

    that being Said, it is now easier to understand why Volvo has already signed an agreement with Ericsson to equip its cars data connections capable of transmitting streaming video with HD quality. The important thing was not that we could watch our favorite series in high-definition, but our cars, in the near future, be able to transmit and receive a vast amount of data.

    Source: NVIDIA
    In motor: