The comment regarding humans not needing LiDAR misses the point, which is that either every other self-driving car group is incorrectly deploying an expensive hardware solution or there are valid technical reasons why that might be a better path to L5.
Implementing L5 without Lidar must be possible. Humans are absolute proof of that. That's not to say Lidar isn't useful, or cost effective given the complexity of the problem, or even just a really good idea that will make autonomous cars better drivers than people. It might be all of those things.
All my comment says is that Lidar is not a requirement for driving, and we have a very obvious proof of that.
Anyone saying that Lidar is necessary for driving is ignoring the hundreds of millions of examples of driving with just a pair of 2d cameras and a highly specialised organic supercomputer.
EDIT: The point another poster made about weather and bright sunshine is a reasonable one - humans can't drive when they can't see either, so Lidar would potentially make an autonomous car better at driving than people. That would be an example of how Lidar is useful, but still not necessary.
Humans also have a organic supercomputer in their brain. Variety of sensors is probably the best approach. I've grown up and seen technology prices on all sorts of weird technology fall (3d printers, sensors, midi controllers). At some point, we will basically give up on perfection on LIDAR and just do cheap multi-LIDAR arrays where the error rate is overcome by just having multiple units.
One could also argue that humans are not proof of anything, as humans frequently make mistakes and conduct dangerous maneuvers whilst driving. There are many road accidents occurring right now, today, around the world, as a result of humans being insufficient drivers.
We can say that something better than human-level driving ought to be possible without LIDAR, though; take the perception of a human (cameras) and replace the brain with a computer that can make better calculations much faster than a human can.
Implementing L5 without Lidar must be possible. Humans are absolute proof of that. That's not to say Lidar isn't useful, or cost effective given the complexity of the problem, or even just a really good idea that will make autonomous cars better drivers than people. It might be all of those things. All my comment says is that Lidar is not a requirement for driving, and we have a very obvious proof of that.
Anyone saying that Lidar is necessary for driving is ignoring the hundreds of millions of examples of driving with just a pair of 2d cameras and a highly specialised organic supercomputer.
EDIT: The point another poster made about weather and bright sunshine is a reasonable one - humans can't drive when they can't see either, so Lidar would potentially make an autonomous car better at driving than people. That would be an example of how Lidar is useful, but still not necessary.