SLAM isn't being feed into machine learning. And when you say "machine learning" I assume you are talking about a model so even in this case is static and DOESN'T change so there is no learning.
SLAM has been here for decades, its really really really old. there haven't been any modifications made other than having it run faster and us having mor powerful phones.
You don't need machine learning to do SLAM and you certainly don't need machine learning to build maps of SLAM interiors.
You need to look up SLAM (Simultaneous localization and mapping).
Mapping data IS used for actuating. Its a big part of Mercedes drive pilot improvements to their new software coming later this year (their using here maps) and GM supercruise.
"The combination of real-time data with precision mapping also improves vehicle control through curves and hills."
Its literally what the map is there for.
I know all about mobileye and there is no fleet learning in REM.
Now if you call it fleet mapping then i can roll with you on that. but there is NO fleet learning. your car doesn't learn. your car maps. and in Tesla Ap1 case it only mapped GPS location making it useless. i don't understand which part of that you can't comprehend?
This is why i love mobileye, they will tell it to you straight. they don't have time to scratch itching ears waiting desperately to hear the next hype.
the machine learning part of google's VPS is most likely doing object recognition. for example it will recognize that this aisle has toothpaste...etc They are not going in and manually plugging in, "over here is lotion and over here is cereal, etc". This way it can constantly be up to date since stores change their layouts all the time.
Last edited:
