I know that Lidar is used by most of the big players, but since it's so expensive there must be good reasons why they're opting for Lidar sensors rather than less expensive visual light cameras and sensors. Is it just that computer vision technology isn't sufficiently advanced?
For example, you can bounce it off the ground under the car in front of you to "see" and measure how fast the car in front of that car is moving. That can be used to anticipate collision risks that even the car in front of you hasn't seen yet (and you cannot see with your own eyes).
Yes, mainly because CV just isn't there yet. Also because Lidar is better suited for the task than a camera+sufficient CV would be. The cost of Lidar is almost inconsequential anyways when compared with the total amount of building, developing, and testing an autonomous vehicle.
You are not always in an area where visible light is sufficient, additionally light sensors are prone to noise related error. If all of your sensors are subject to the same errors they may collaborate on mistakes
