3 ms·
Not sure if this is possible, but couldn't they capture all the LIDAR inputs from a car being driven by a person, then feed that into the software to see if it
by refurb 8y ago
Not sure if this is possible, but couldn't they capture all the LIDAR inputs from a car being driven by a person, then feed that into the software to see if it actually recognizing objects correctly?
Or even do it on a closed track. That way you could throw all sorts of things at it like kids, bikes, garbage cans, whatever.
Should be able to see what works and what doesn't work pretty quickly.
EDIT: Never mind! That's what Waymo does in the article another person posted.
https://www.theatlantic.com/technology/archive/2017/08/inside-waymos-secret-testing-and-simulation-facilities/537648/ https://www.theatlantic.com/technology/archive/2017/08/insid...
- Klathmon 8y agoWhen it comes to "closed track" and other "simulation" style things, you need to be very careful when ML is involved. I don't know if Uber/Waymo/Others have "solved" this problem, but ML systems have a tendency to "learn" the wrong thing. You could run millions of simulations (both virtual or on a closed course) with "test obstacles", but the second the car encounters one a child with red balloons in the early morning on a foggy day it might not know what to do with it because that exact scenario wasn't in any simulations.
- caboteria 8y agoThey'd be using the simulator for validation, not training.
- refurb 8y agoExactly. Throw something at it the system should stop for (e.g. a child). If it doesn't, go back to the drawing board and figure out why. Agreed the test track is not ideal, that's why I was wondering if they could throw real recorded data at it. No risk to the public, but puts the system through real-world trials.