http://www.caltech.edu/news/new-camera-chip-provides-superfine-3-d-resolution-46425 LIDAR on a single chip looks like it is going to be 300-400 a unit retail. http://spectrum.ieee.org/cars-that-think/transportation/sensors/cheap-lidar-for-automatic-braking edit 100 a unit. that reporter is even more of a gonzo.
lol. http://www.subaru.com/engineering/eyesight.html The author has a full resume of not liking Google and automated cars. Adding him to the counter-point file.LIDAR, a sensor most in the field view as essential to giving a car the gift of sight, can cost more than twice as much as the vehicle itself. A popular early model from Velodyne, used by many, costs $75,000.
Not at all the same thing. Subaru has an entirely different system with an entirely different purpose: assistance instead of full control, meaning it can do with a cheap sub-$1000 intelligent camera. The kind of LIDAR that Google needed -the one on their Lexuses and Priuses - was the $75k Velodyne. They're now working with a sub-10k model, I think. While I don't think it is a hurdle to the development of self-driving cars, it is definitely necessary to develop better and cheaper models. There's a reason they haven't announced the specs of the budget lasers yet because that's when people will likely become sceptical. You really do need a high level of precision, quality assurance and speed for full control, none of which is cheap at the moment but inevitably will be.
The interesting thing is that reading about the LIDAR-on-a-chip is how do thousands of cars keep who's radar is who's? You'd have to put a serial number in the beam somehow, which means that you can put a detector on the street and track vehicles very easily. 10 years is three machine generations. I have no doubt that by the time these cars are ready, the production demands will drop the price per unit much like CPU prices have plunged. If they can make a $5 Raspberry Pi, they can find a way to do cheap LIDAR using the steel/aluminum in the car as the antenna. The thing is that what seems improbable now only needs time talent and money to make commonplace. In 12-14 years when it is time for me to buy a new car, I don't doubt that this tech will be an option.