As I remember it, the driver got a lot of heat for fumbling with her phone (or 2nd computer?) right before the accident. I don't think however that 1.2s is a bad reaction time for a complex situation.
Would it have killed the developers to make the car sound its horn when it gets into this absurd 1s "action suppression" mode?
> the driver got a lot of heat for fumbling with her phone (or 2nd computer?) right before the accident
Based on news stories I found, she was glancing at a television show on her phone [1].
> make the car sound its horn when it gets into this absurd 1s "action suppression" mode?
If they added the suppression because there were too many false positives, that would just have resulted in the car honking at apparently arbitrary times. It's just converting the garbage signal from one form into another. It's still too noisy to be reliable.
1: https://www.azcentral.com/story/news/local/tempe/2019/03/17/...Vasquez looked down 166 times when the vehicle was in motion, not including times she appeared to be checking gauges [...] In all, they concluded, Vasquez traveled about 3.67 miles total while looking away from the road. [...] starting at 9:16, Vasquez played an episode of “The Voice,” The Blind Auditions, Part 5, on her phone.
> If they added the suppression because there were too many false positives, that would just have resulted in the car honking at apparently arbitrary times. It's just converting the garbage signal from one form into another. It's still too noisy to be reliable.
I love how they went from "our vision system is too unreliable to have warning signals every time it doesn't know what's in front of it" to "okay let's do it anyway but just not have warning signals". Like it didn't make them stop and think "well maybe basing a self-driving car off of this isn't a good idea".
I don't know why you're downvoted, because I find your comment funny. Reminds me of a similar real fix for a race condition I found somewhere in one of the companies I worked before:
Compared to classifying a pedestrian with a bicycle crossing the street in the dark it is easy to track the gaze of the safety driver and stop the vehicle when they are not looking at the road.
> I don't think however that 1.2s is a bad reaction time for a complex situation.
1.2 seconds after hitting a predestrian. That's a pretty poor reaction time. Typically you want to apply the breaks before you come into contact with a person.
That has nothing to do with the reaction time. Yeah, they weren't looking at the road, but that's a separate issue that doesn't involve reaction time. Going from an unexpected buzzer to full control of the vehicle in a single second is a pretty good speed.
No it's not, and hitting the brakes isn't "full control". Honestly, how long do you think it would take you to hit the breaks if something jumped out at you? It would be almost instant.
Edit: ok, I was just making a bit of a joke at first, but I looked it up. Reaction times vary by person, but tend to be between 0.75 and 3 seconds. 2.5s is used as a standard, so I guess I have to conceded that 1.25s is pretty good... I guess, for whatever that's worth.
Just a few minutes ago I had a very similar situation, although at day time. I was going at about 40 kmph straight forward my lane near a bus standing at a bus stop at the right side of the street. At the moment I was passing the bus, a young person almost jumped out in front of me, maybe 15-25 meters away. They were hidden behind the bus, so I had no way to see this coming in advance. Fortunately they realized I was coming and backed off and I also managed to stop completely before them.
So if my reaction was 5.7 seconds, I'd definitely have applied brakes far too late. I conclude the total time from classifying the object moving into my way and applying brakes was less than a second. (And btw, my car has emergency braking / pedestrian avoiding system and it didn't trigger, so I was faster).
>Edit: ok, I was just making a bit of a joke at first, but I looked it up. Reaction times vary by person, but tend to be between 0.75 and 3 seconds. 2.5s is used as a standard, so I guess I have to conceded that 1.25s is pretty good... I guess, for whatever that's worth.
We have a rare sight here: someone not being right, learning more about the situation, changing their opinion, and then making an edit about it all.
Kudos, and thanks for making this a better place for discussions :)
Would it have killed the developers to make the car sound its horn when it gets into this absurd 1s "action suppression" mode?