4 ms·
Do you not feel that their relative success during your ride in Chandler is at least partially due to the specific location (easy suburban streets) and the geof
by cycrutchfield 5y ago
Do you not feel that their relative success during your ride in Chandler is at least partially due to the specific location (easy suburban streets) and the geofenced boundaries (they require high quality ground truth labels in order to be successful). If we took the car and dropped it in the middle of Manhattan would it work as well?
- ra7 5y agoIt’s easy to say Chandler is just “easy suburban streets”, but for a computer to drive by itself it’s completely non-trivial. In any case, they are already giving private beta rides in SF, so you will soon see how they tackle dense urban streets. They recently started mapping NYC, so I assume that’s coming in the near future too. Waymo (and others) are very clear that geofencing and prior knowledge of roads are necessary for safe operations. Their safety standards are extremely high because there’s no driver for fallback — the vehicle is responsible end-to-end. Geofencing allows them to validate extensively and that means when they say it works, it really does. I think the mapping argument is pretty overblown at this point. Most of them have figured out how to do it efficiently and keep it up-to-date.
- cycrutchfield 5y agoIt’s more a question of scale (can they afford to map the entire US, or the entire globe?), and of over-reliance on features which can become stale.
- ra7 5y agoThey've claimed they can map efficiently and the maps are self-updating from the fleet [1]. Waymo and other SDCs have said in the past that mapping isn't as big of an issue, but we will have to wait and see. [1] https://blog.waymo.com/2020/09/the-waymo-driver-handbook-mapping.html https://blog.waymo.com/2020/09/the-waymo-driver-handbook-map...