In the piece — titled “Can You Fool a Self Driving Car?” — Rober found that a Tesla car on Autopilot was fooled by a Wile E. Coyote-style wall painted to look like the road ahead of it, with the electric vehicle plowing right through it instead of stopping.
The footage was damning enough, with slow-motion clips showing the car not only crashing through the styrofoam wall but also a mannequin of a child. The Tesla was also fooled by simulated rain and fog.
Because Musk insists that cameras are better and that LiDAR is flawed
That’s not really true.
He use lidar in SpaceX because he knows it’s the right tool for their specific job.
His stance is it’s not that cameras are better, but that cameras have to be so good for a truly AV that putting effort into both means you’re not going to make your cameras good enough to do it and rely on lidar instead. That and cost.
If the car can’t process and understand the world via cameras, it’s doomed to fail at a mass scale anyway.
It might be a wrong stance, but it’s not that lidar is flawed.
Tesla even uses lidar to ground truth their cameras
Edit: just adding a late example - Waymo, Cruise, and probably everyone out there still use humans to tell the car what to do if it gets stuck. I even bet Tesla will if they ever launch a robotaxi as they need a way to somehow help the car if it gets stuck. When we see these failures with Waymo and Cruise, it’s less “is something there” and more “I don’t understand this situation”. The understanding comes from vision. Lidar just gives the something is there, but it isn’t solving their problem.
I think the bigger issue is that he is saying redundancy is not important. He thinks cameras could be good enough, well fine, but the failure results in loss of life so build in redundancy: lidar, radar, anything to failover. The fact that cutting costs OR having a belief that one system is good enough is despicable.
Ya, no redundancy is a problem for sure.