4 ms·
We have working self-driving cars today, and they need no special infrastructure. Detecting painted lane markings and other vehicles are the easy parts of the d
by Daniel_Newby 13y ago
We have working self-driving cars today, and they need no special infrastructure. Detecting painted lane markings and other vehicles are the easy parts of the data processing system.
Furthermore, special infrastructure is unacceptably vulnerable to deliberate disruption. A system that depended on radio communication could be gridlocked with a modified $50 microwave oven.
- Siecje 13y agoWhat if you can't detect the painted lines on the road, because of snow?
- IanCal 13y agoJust like humans, these cars will look at multiple things. If they can see the lines then it helps, if not that doesn't mean they can't drive. They know the road and where they are, they can see what's around them and how fast they're travelling. For humans to beat machines at driving, we need to consider situations that require complex reasoning that humans are good at. Essentially, for any "But what would they do if X" you need to say what humans need to do, and why it's a complex task.
- w-m 13y agoThere's many other indicators like reflectors, the snow condition, trails from other cars, snow height, ... Whatever you can use to find out where the road is supposed to be, the self-driving car can use as well.
- Daniel_Newby 13y agoThe other responses are good. There will also probably be databases of landmarks like buildings and light poles that can be used to make a centimeter-accurate visual positioning system.