Detecting what’s around is the first crucial task of an autonomous vehicle (the second one is to understand what’s there and the third is foreseeing the likely movements in order to take appropriate decisions).
By far the best technology for detecting (still and moving) ambient objects is LIDAR (LIght Detection And Ranging). To get a feeling on how it works and its evolution have a look at “a brief history of LIDAR” by BCC Research.
The problem with LIDAR is the cost. Best in class equipment may cost up to 75,000$ (Velodyne) and even more recent products (like Waymo) have a tag price of 7,500$, way too much for a mass market use.
This is why semi-autonomous vehicles, like Tesla, rely on different technologies for ambient detection. Unfortunately these technologies (image recognition based on digital cameras and radar) are not as accurate as LIDAR. A LIDAR works well in sunlight as in the dark (not being constrained by light conditions as a digital camera is) and works well in foggy and rainy conditions (watch the clip).
Now Luminar is aiming at what they see as a third LIDAR generation with a target price of 500$ in a form factor of a can of coke (see photo).
It is a car grade equipment, able to stand a broad range of temperature over several years and hundreds of thousand of kilometres in a bumping vehicle, drawing just 15W of power and able to scan the ambient up to 250 meters away with a precision of 1cm.
The Luminar LIDAR has a limit in the range of vision, it covers only what’s in front of it, 120° of range, making it suitable for highway driving but not for urban environment where a 360° field of view is required.
Luminar is expecting to deploy its system on many test cars, on six lane highways first and use the data to refine the software (remember, detection is the first step, than you have to understand what is being detected and for this software is needed). The aim is to have commercial equipment available for cars manufactured in 2022/2023.