We already have cars today that can stop a human from crashing - there are cars that will automatically apply the brakes to avoid a collision (Volvo has one for pedestrians, Toyota for objects, some are in production), and can maintain a set distance to the car in front of them on a roadway (auto cruise control).
Those are both a form of automation / robotics, even if the car makers don't like to call it that.
The human eye is an amazing sensor, one we can't even get close to with our current technology.
The human eye is amazing, yes, but it has severe limitations relative to other types of sensors, which we have to compensate for with systems like headlights and street lamps. It's also backed by a recognition system that wasn't really ever designed to work at >20 MPH.
It's a pretty bad sensor. We can only see certain wavelengths need corrective devices, have a blind spot and can only focus on a small part of our visual field.
And if the system gets tired and the sensors shut down.
At night: much better than a human who can only see visible light
Stop for a dog: no problem at all
You haven't addressed the real problems:
In the snow: lane markers disappear
False positives: a load of snow falling off the car in front of you, or a garbage bag flying across the road, can both cause the car to go paranoid.
It's still vastly better than ugly bags of mostly water
Before robotic cars are viable, we'll first need cars that prevent human drivers from causing an accident. If a car can't stop a human driver from crashing, how can it stop itself?
And if people were to see accident rates plummeting for new cars, they would have the confidence that robotic vehicles will be safe too.
We can test robotic safety before the first robotic car takes to the road.