AP911:

For autonomous driving to work properly, it needs redundancy so that if one method didn't work well in a certain situation another method should. I think that is what Mercedes are doing for example when there is fog and the camera is useless in determining what is in the front of the car. In addition, the recent advances in Machine learning and AI from deep learning techniques will revolutionize the way autonomous driving work in the same way it revolutionized other AI domains. The reason for this is that deep learning utilizes neural networks to replicate human intelligence and not the current autonomous drive systems that work by following rules. And deep learning based autonomous drive will learn from the environment of the system, which means for different countries the same system will adapt itself to the traffic rules in each county. However, for deep learning to work, a very large data set is required which means long hours of testing and learning.

The advances in autonomous driving will come from outside the automotive industry. Look at CES2018 and you will see how many companies and researchers are working on this problem. It will not take long for level 3 autonomous drive to become commonly used. Audi has beaten Tesla to it in the new A8, however, its level 3 is only limited to 60km/h. It also the first car to use LIDAR which helps greatly in detecting surrounding objects and depth. 

What many might not know is that Tesla's self driving capabilities are actually worst that they were 3 years ago, because they no longer use the Mobileye systems in their newer cars. 

 

Yeah, that redundant system is called human driver. Smiley


--