street view images so we can navigate ourselves, but for a vehicle, autonomousi vehicle, this needs to be done at a much higher precision. we need centimetre level- detailed maps of entire cities, which is actually a remarkably solvable problem but very - expensive to do and even worse to maintain. - there is another limitation with current approaches, i in terms of the sensors thati get used are very expensive sensors, something called lidar which is a way of sending out. beams of laser light and measuring the time for that light to reflect back and that gives you a sense of how fari things are away. that s a wonderful sensor,
at a much higher precision. we need centimetre level- detailed maps of entire cities. which is actually a remarkably solvable problem but very - expensive to do and even worse to maintain. - there is another limitation with current approaches i in terms of the sensors - they get used a very expensive sensors, something called lidar which is a way of sending out. beams of laser light and measuring the time for that light to reflect back and that gives you a sense of how fari things are away. it s a wonderful sensor, i i love depth information, i used it in my previous| work but it is expensive and difficult to - integrate in a vehicle. the third one is how- the intelligence for these is setup. today, most approaches - to autonomous driving are based on a very modular system where you sort of first try| and detect all the cars