Uber Opening Robotics Research Facility in Pittsburgh to Build Self-Driving Cars

Re:Uber Opening Robotics Research Facility in Pittsburgh to ... by ForHackernews on Tuesday February 03, 2015 @11:44AM (#8991375)

(Leaving aside privacy issues for the moment) An existing car manufacturer could build sensors/data collection into their entire fleet over a period of several years. They could monitor the environment around millions of cars and learn to model driver behavior.

I don't know how many Google Maps cars Google has, but if car manufacturers got serious about data collection, there's no way Google could compete with the scale of their installed user base.

[ Parent | on HN ]

Re:Uber Opening Robotics Research Facility in Pittsburgh to ... by TeMPOraL on Tuesday February 03, 2015 @11:59AM (#8991466)

That's just collection, but you need also need real-time processing. I'm pretty sure Google has more software engineers and computer scientists at hand than any car manufacturer.

[ Parent | on HN ]

Re:Uber Opening Robotics Research Facility in Pittsburgh to ... by ForHackernews on Tuesday February 03, 2015 @12:09PM (#8991546)

Sure, but a car company could hire software people, or contract with somebody like IBM.

Most of what Google is doing with their self-driving car is not a difficult software or AI problem, they've simply built extremely, extremely detailed maps of a limited subset of roads: http://www.wired.com/2014/05/google-self-driving-car-maps/

Contrast that with the DARPA challenges, where a car needed to navigate unpredictable off-road terrain: https://en.wikipedia.org/wiki/DARPA_Grand_Challenge

Doing it Google's way, you need some sensors and onboard processing, but the amount of hardware is actually pretty minimal compared to Carnegie Mellon's self-driving cars. Most of the fancy hardware is for making the maps, not driving the car.

[ Parent | on HN ]