Danny Shapiro, Senior Director Automotive, Nvidia, confirmed that Drive PX comes out in May for $10,000, something that CEO Jen-Hsun Huang said at the GPU technology conference keynote.
Danny also added that Google and Audi are among several companies using Nvidia technology and without having Nvidia to confirm, Elon Musk’s Tesla cars is also working with Nvidia on to develop a self-driving car.
Nvidia is confident that they are the leader in this market segment and that they peaked interest from a lot of automakers. Their Drive PX can handle 12 inputs from 12 cameras, use an ultrasonic sensor array as well as LiDar (Radar that shoots light rays that you might know from the Google Car prototype).
Danny also explained that future cars might use sensor data to improve self-driving based on predefined conditions, as well as data that the car computer would get from cameras and sensors. The future car should combine multiple different data sources, including camera inputs, LiDar, ultrasonic sensors and other types of sensors.
This is also something that Elon Musk confirmed in talks with Nvidia’s CEO. Musk even thinks that self-driving is virtually a solved problem, despite the fact that it will take years before we see commercial self-driving cars on the road.
Nvidia thinks that deep learning, a buzzword of this tech conference, will bring a new factor to play. With deep learning Nvidia hopes that your car will know the difference between a Fedex truck and an ambulance. The collected data can have a higher degree of accuracy than data collected and processed by humans today, of course if everything is done properly.
We saw a demo of a Camaro car body packed with multiple cameras connected to Nvidia Drive PX. These cameras were tracking what was going on around the car, showing us a glimpse of the future. These wide angle cameras would be positioned all over the the car and they would feed the Drive PX system with all necessary information. There is a connection for ultrasonic sensors and LiDar as well, and we saw a demo with multiple wide-angle cameras that would get the data that should help the car computer to get closer to true self-driving.
Processed in the right way, such streams of data could be a great starting point for an autopilot option in future cars. Car manufacturers are working on it, but this is still a long way from being shipped in actual cars. Who knows, maybe the Tesla X that got delayed to Q3 2015 might be the first car with limited self-driving capabilities?