I know Argo is old news but this just came out, more at his attempt as to "why". Not sure I agree.
https://www.hotcars.com/ford-multibillion-dollar-self-driving-investment-complete-failure/
I think I would agree that Argo was shut down more for business reasons than tech reasons. Juergen Reers, a senior managing director at consulting firm
Accenture says that the tech is advanced enough for L4 but is still very costly:
"Overall, the level of technology is well advanced, but a key challenge is the cost. In high levels of automated driving, a high degree of redundancy is needed. Combining camera, radar and lidar systems with high-definition maps and high computing requirements would make a Level 4 vehicle too costly for individual use. So we expect these vehicles to be used as shuttles, carrying multiple passengers, ideally 24/7 to absorb the cost."
And Raj Rajkumar, a professor in the electrical and computer engineering department at Carnegie Mellon University, says that a vision-only approach is cheap enough but is not good enough yet for reliable L4:
"First, some companies only want to rely on cameras for economic reasons—making the final system very affordable thanks to the fact that cameras in smartphones have made them ubiquitous, compact and inexpensive. Unfortunately, today’s vision technology using only cameras won’t match, for a long time to come, the cognitive capabilities of human eyes plus human neural processing. Two, lidar is a critical component, but while costs are coming down, the costs are still on the higher side. Radar, ultrasonic and GPS are well within acceptable cost limits already."
Source:
When Will Cars Be Fully Self-Driving?
So I think the big issue is cost and the business side more than tech. That is not to say that the tech is solved yet but the tech needed for reliable L4 exists, it is just expensive. In theory, the robotaxi model should be profitable if you can bring down costs and deploy safe, reliable and convenient ride-hailing in high demand areas like cities. But that requires a lot of things to come together.
Obviously, the ultimate goal is to achieve safe, reliable L4 that is also cheap enough. When that happens, we will likely see mass scale of fully autonomous cars. One approach is to start with expensive tech that is good enough for reliable L4 and then try to bring costs down (Waymo). We are seeing cost of lidar already come down. So I imagine that costs will eventually be low enough. So I think the Waymo approach will work, it is just a matter of when and can it happen soon enough. The other approach is to start with low cost and then try to make the tech good enough for reliable L4 (Tesla). My guess is the two approaches will eventually merge. So we will see Waymo-like software but combined with fewer/cheaper sensors. For example, Mobileye says that their "eyes off" system will use surround vision, surround radar (with their in-house cheaper radar) and only 1 front lidar. So the idea is to find that middle ground where you still have some extra radar and lidar for added reliability but still keep costs reasonable.
electrek.co
