Printable View
The cyclist was detected 6 seconds before impact as an object, it was 1.3 seconds before impact the system worked out it was a cyclist and deemed emergency braking was nessesary. A very damning report from the NTSC.
Full preliminary report - https://www.ntsb.gov/investigations/...010-prelim.pdf
Dave sums it up well in his video below.
https://youtu.be/DbT2aBlC0no
I wonder if a human driver would have reacted any faster.
A human driver wouldn’t have seen the object 6 seconds out like the car did in the dark.Quote:
Originally Posted by DiscoMick
Six seconds is a long time and IMHO a human would have immediately seen the danger.
It calls into question the whole hype about autonomous vehicles.
Did the automatic braking give so many false positives that it was disabled?
This question of discrimination and decision making is fundamental to the success of truly autonomous vehicles , without a driver who de facto in this case seems to be being blamed because she was monitoring the autonomy.
The true advantage of autonomous vehicles is that they are supposed to operate without a driver and it is the potential cost savings that have bus operators and taxi operators salivating.
Regards Philip A
It was dark either side of the road, a human wouldn't have seen the cyclist 6 seconds out, not a chance. Autonomy is still very much in its infancy, it will come, but I think it will be decades, not years.
A few people have posted footage from the same spot, but using their own cameras, but it's a lot lighter than what is shown on the crash video. One is below.
YouTube
Yeah, but cameras never give an accurate representation of light levels. I’ve got some in car footage of a back street near my house which is very dark with infrequent street lighting but the camera makes it look well lit.
Have a look at the accident statistics. They don't look too good? Just have a think about how many driver decisions are made daily that do not result in an accident compared to the ones that do have that result.
Regardless of what reasonable assumptions are made in the calculation, the result shows that for an autonomous vehicle to have a safety record even as good as the average driver, requires software reliability that has yet to be achieved in any remotely comparable software application. Where this sort of reliability is really needed, such as in aircraft fly-by-wire systems, typically systems are triplicated, using separately developed software and 'majority rule". And this sort of existing application is far simpler than autonomous driving.
While I think autonomous vehicles will come, I agree it will be decades rather than years.
More people die from cancer than on the roads. Most drivers do well, I think. I question if auto driving will produce the improvements that are predicted in the unpredictable road environment. But I can see semi-auto buses and trains, and maybe even planes, doing well.
Autonomous trains already operate in a number of places, but you need to bear in mind that these are in a controlled environment, with restricted access to the tracks. Aeroplanes also operate autonomously in some cases - in a controlled environment, access only to aircraft with a clearance, and with their operation closely monitored by an on-board pilot as well as air traffic control. Mine and warehouse vehicles operate autonomously in some cases, but again, these are in a controlled environment with no public access.
All these have to deal with a far simpler environment than does an autonomous vehicle. I think even buses are a long way off.