Uber Self-Driving Car and the LIDAR
Recently I’ve read a post “It certainly looks bad for Uber” on Brad Ideas blog. And there are some points that I can’t agree with.
Brad made very good analytics, but there are some simplifications that may mislead less educated audience. I’m not an expert btw, but I’ll try to explain.
Brad points that the LIDAR system that was switched off while car was training a camera+radar systems. Brad said, “the LIDAR should remain on as a backup, triggering braking in exactly this sort of situation.” Unfortunately, everything is not that simple.
No doubts, it would be the best outcome if the backup system would prevent that tragic accident.
How to give control
Maybe it is not even possible to make a backup system there. Or maybe it was too expensive or too long to implement. The main reason: these systems are working by collecting a data from a lot of heterogeneous sources, and then sending all the data to a resulting decision.
And those sources may have a lot of cross-effects of different orders.
Therefore, it is needed to have two instances of compensative brains on board. Where everything is connected to the first one, and everything but a LIDAR is connected to the second one.
And after all it is very important to understand in what moment to give control to the full fledged brain.
It seems very fucking difficult, to be honest. As always, one have to stop somewhere. Even for a company as Uber. Bad luck.