[home][rumors and news][model release matrix][dealer network][desktop calendar][exhaust notes][tov forums][links][search][sponsors][garage][login]

Tire Rack Upgrade Garage
 Search for a Dealer:
 Canadian Flag US Flag
 Honda Acura
 ZIP  
NSX GT3 Evo To Compete Globally in 2019
More.......................
2019 Acura ILX Arrives with Dynamic New Styling, Major Technology Upgrades and New A-Spec Treatment
More.......................
Honda Joins with Cruise and General Motors to Build New Autonomous Vehicle
More.......................
Honda reports September 2018 sales numbers
More.......................
Honda recalling 232,000 2018 Accords and 2019 Insights for rear camera software update
More.......................
2019 Accord arriving in showrooms by November
More.......................
Is a base trim of the Civic Type R finally arriving for MY 2019? (More MY2019 info inside)
More.......................
Team Honda Research wins ST2 at NASA Nationals with their NC1 NSX
More.......................
Type R --> Re: Top Gear: Here's what seven months living with a Honda Civic Type R is like
Join Discussion......
Today's Reading Links --> Re: Hydrogen vs Battery-tanks
Join Discussion......
General Talk --> Re: So they took apart a Tesla 3 and found.......
Join Discussion......
Accord --> Re: Accord 2.0 Touring: First Impressions
Join Discussion......
Professional Motorsports --> Re: F1 - 2018 United States Grand Prix Race Spoilers
Join Discussion......
IMA/Hybrids --> Re: Check IMA System light is on - what could it mean?
Join Discussion......
Today's Reading Links --> New status symbol: Dream garages
Join Discussion......
Ridgeline - General Talk --> Re: Honda to bring mystery off-road truck to 2018 SEMA
Join Discussion......
Today's Reading Links --> Re: European CO2 madness and why electromobility is pointless
Join Discussion......
General Talk --> Re: CR-V Hybrid 4wd iMMD
Join Discussion......
General Talk --> Re: The Dirt on Clean Electric Cars
Join Discussion......
RDX --> Re: Drove one over the weekend...
Join Discussion......
Fuel Cell Technology --> Re: DMV Stickers, we don't need no stinking stickers
Join Discussion......
S2000 --> Bridgewater Acura Meet
Join Discussion......
Professional Motorsports --> Re: 2018 MotoGP
Join Discussion......
First Drive: 2019 Honda Pilot
Read Article....................
2019 Honda Pilot PR Photo Gallery
Read Article....................
First Drive: 2019 Honda Insight
Read Article....................
2019 Honda Insight PR Photo Gallery
Read Article....................
First Drive: 2019 Acura RDX
Read Article....................
2019 Acura RDX Features & Specifications
Read Article....................

[fancy] [flat] [simple]
TOV Forums > General Talk > > Re: Mobileye claims their system detected the pedestrian, even with the crappy Uber video as input

Go to:

Viewing Threshold (What is this?)

Thread Page - 1 2 3 4 [5]
Author
  Post New Thread
notyper
Profile for notyper
Re: Mobileye claims their system detected the pedestrian, even with the crappy Uber video as input    (Score: 1, Normal) 03-27-2018 14:14
Reply to This Message Attach Quote to Reply
JeffX wrote:
fishchan wrote:
JeffX wrote:
Mobileye claims their system identified the pedestrian and her bicycle about 1 second before impact, using the terrible quality video that Uber released as the input to their system.




But that is still too late. Did they say it can avoid the accident in the 1 sscond time if it is driving the car?



1) 1 second of braking from 40mph is plenty of time to slow the car to a non fatal impact, or possibly no impact.
2) It seems that video was nowhere close to representative of what a typical sensor suite (including cameras) would have detected, so a properly functioning system should've identified the pending collision much sooner than 1 second before impact.



Yep, 40 mph is about 59 ft/s. Any modern car can pull close to 1 g of decel at low speeds in dry conditions. So assuming near instantaneous computer reactions I would think they should have been able to cut speed in half before impact. That's a 75% reduction in impact energy and would increase survival odds immensely.

SC

Hondu
Profile for Hondu
Re: Mobileye claims their system detected the pedestrian, even with the crappy Uber video as input    (Score: 1, Normal) 05-24-2018 12:21
Reply to This Message Attach Quote to Reply
notyper wrote:
JeffX wrote:
fishchan wrote:
JeffX wrote:
Mobileye claims their system identified the pedestrian and her bicycle about 1 second before impact, using the terrible quality video that Uber released as the input to their system.



But that is still too late. Did they say it can avoid the accident in the 1 sscond time if it is driving the car?



1) 1 second of braking from 40mph is plenty of time to slow the car to a non fatal impact, or possibly no impact.
2) It seems that video was nowhere close to representative of what a typical sensor suite (including cameras) would have detected, so a properly functioning system should've identified the pending collision much sooner than 1 second before impact.



Yep, 40 mph is about 59 ft/s. Any modern car can pull close to 1 g of decel at low speeds in dry conditions. So assuming near instantaneous computer reactions I would think they should have been able to cut speed in half before impact. That's a 75% reduction in impact energy and would increase survival odds immensely.

SC




NTSB has issued a preliminary report.

http://online.wsj.com/public/resources/documents/NTSBuber.pdf?mod=article_inline

superchg2
Profile for superchg2
Re: Mobileye claims their system detected the pedestrian, even with the crappy Uber video as input    (Score: 1, Normal) 05-24-2018 12:53
Reply to This Message Attach Quote to Reply
Hondu wrote:
notyper wrote:
JeffX wrote:
fishchan wrote:
JeffX wrote:
Mobileye claims their system identified the pedestrian and her bicycle about 1 second before impact, using the terrible quality video that Uber released as the input to their system.



But that is still too late. Did they say it can avoid the accident in the 1 sscond time if it is driving the car?



1) 1 second of braking from 40mph is plenty of time to slow the car to a non fatal impact, or possibly no impact.
2) It seems that video was nowhere close to representative of what a typical sensor suite (including cameras) would have detected, so a properly functioning system should've identified the pending collision much sooner than 1 second before impact.



Yep, 40 mph is about 59 ft/s. Any modern car can pull close to 1 g of decel at low speeds in dry conditions. So assuming near instantaneous computer reactions I would think they should have been able to cut speed in half before impact. That's a 75% reduction in impact energy and would increase survival odds immensely.

SC




NTSB has issued a preliminary report.

http://online.wsj.com/public/resources/documents/NTSBuber.pdf?mod=article_inline


"According to Uber, emergency braking maneuvers are
not enabled while the vehicle is under computer control, to
reduce the potential for erratic vehicle behavior.

The vehicle operator is relied on to intervene and take action.

The system is not designed to alert the operator.
"

WTF?

TonyEX
Profile for TonyEX
Re: Mobileye claims their system detected the pedestrian, even with the crappy Uber video as input    (Score: 1, Normal) 05-24-2018 13:30
Reply to This Message Attach Quote to Reply
superchg2 wrote:
...
"According to Uber, emergency braking maneuvers are
not enabled while the vehicle is under computer control, to
reduce the potential for erratic vehicle behavior.

The vehicle operator is relied on to intervene and take action.

The system is not designed to alert the operator.
"

WTF?



Look, it's like cruise control...

THE DRIVER IS EXPECTED TO PAY ATTENTION!!

So pretty much, unfortunately for Tesla and autonomous vehicles: In general there are two kinds of drivers:

(1) Drivers that love to drive, they want to drive.
(2) Drivers that love to be a passenger, they really don't want to drive.

IMHO, it's the (2) above that use all of those nannies. And they don't really want to drive, so they don't pay attention. 99% of the time, it may work: ACC, lane keeping, seat massagers, smart phone connectivity, you name it. But that 1% of the time, when they should be looking down the road, they are instead texting... and WHAM!

Now, the (1) above, the ones that most likely would be engaged with the driving and could be expected to react on that 1% of the time.. well, heck, they turn OFF the nannies and drive the car themselves. Precisely because these systems are so intrusive and issue too many false warnings. (*)

BTW, 1 second warning is hard to react. That pedestrian was toast regardless. Didn't the damn pedestrian see the car?

(*) The first time I drove a Honda car on the 405 with the nannys turned on I almost rear ended the car in front of me. As we were slowing down, the thing suddenly FLASHED a warning on the dash and my attention was drawn to that, not to the car ahead. Yikes. I KNOW HOW TO SLOW DOWN... ever since, the first thing I do when I get into a Honda car is to turn OFF the nannies.

notyper
Profile for notyper
Re: Mobileye claims their system detected the pedestrian, even with the crappy Uber video as input    (Score: 1, Normal) 05-24-2018 21:03
Reply to This Message Attach Quote to Reply
superchg2 wrote:

"According to Uber, emergency braking maneuvers are
not enabled while the vehicle is under computer control, to
reduce the potential for erratic vehicle behavior.

The vehicle operator is relied on to intervene and take action.

The system is not designed to alert the operator.
"

WTF?



Yeah, assuming that wasn't a typo, that's pretty f***ed up. You have a system which apparently was able to identify a potential issue many seconds in advance, and was able to actually identify the target as a bicycle well over 1 second in advance.

Yet, the system was not programmed to react to prevent a collision? I know they have a driver on board, but what the hell? If this is true, and the driver was made aware in their training that they were responsible for emergency responses, then the driver is in some trouble here, and because they were on the job, so is the company.

SC

superchg2
Profile for superchg2
Re: Mobileye claims their system detected the pedestrian, even with the crappy Uber video as input    (Score: 1, Normal) 05-24-2018 23:44
Reply to This Message Attach Quote to Reply
notyper wrote:
superchg2 wrote:

"According to Uber, emergency braking maneuvers are
not enabled while the vehicle is under computer control, to
reduce the potential for erratic vehicle behavior.

The vehicle operator is relied on to intervene and take action.

The system is not designed to alert the operator.
"

WTF?



Yeah, assuming that wasn't a typo, that's pretty f***ed up. You have a system which apparently was able to identify a potential issue many seconds in advance, and was able to actually identify the target as a bicycle well over 1 second in advance.

Yet, the system was not programmed to react to prevent a collision? I know they have a driver on board, but what the hell? If this is true, and the driver was made aware in their training that they were responsible for emergency responses, then the driver is in some trouble here, and because they were on the job, so is the company.

SC



At the very least, an alarm should have sounded along with a flashing light as the lady and her bicycle was scanned (six seconds before the crash), which might have given the driver enough time to take preemptive action.

CarPhreakD
Profile for CarPhreakD
Re: Mobileye claims their system detected the pedestrian, even with the crappy Uber video as input    (Score: 1, Normal) 05-25-2018 13:59
Reply to This Message Attach Quote to Reply
Another point- every other OEM uses two people when testing autonomous cars; while Uber for some reason is asking the driver to monitor all the systems as well as perform the emergency braking. It was never going to end well.
superchg2
Profile for superchg2
Re: Mobileye claims their system detected the pedestrian, even with the crappy Uber video as input    (Score: 1, Normal) 05-25-2018 16:43
Reply to This Message Attach Quote to Reply
This fiasco, as well as the death wish accidents with Tesla's, have probably set autonomous cars back 5 years.


owequitit
Profile for owequitit
Re: Mobileye claims their system detected the pedestrian, even with the crappy Uber video as input    (Score: 1, Normal) 05-26-2018 03:11
Reply to This Message Attach Quote to Reply
TonyEX wrote:
superchg2 wrote:
...
"According to Uber, emergency braking maneuvers are
not enabled while the vehicle is under computer control, to
reduce the potential for erratic vehicle behavior.

The vehicle operator is relied on to intervene and take action.

The system is not designed to alert the operator.
"

WTF?



Look, it's like cruise control...

THE DRIVER IS EXPECTED TO PAY ATTENTION!!

So pretty much, unfortunately for Tesla and autonomous vehicles: In general there are two kinds of drivers:

(1) Drivers that love to drive, they want to drive.
(2) Drivers that love to be a passenger, they really don't want to drive.

IMHO, it's the (2) above that use all of those nannies. And they don't really want to drive, so they don't pay attention. 99% of the time, it may work: ACC, lane keeping, seat massagers, smart phone connectivity, you name it. But that 1% of the time, when they should be looking down the road, they are instead texting... and WHAM!

Now, the (1) above, the ones that most likely would be engaged with the driving and could be expected to react on that 1% of the time.. well, heck, they turn OFF the nannies and drive the car themselves. Precisely because these systems are so intrusive and issue too many false warnings. (*)

BTW, 1 second warning is hard to react. That pedestrian was toast regardless. Didn't the damn pedestrian see the car?

(*) The first time I drove a Honda car on the 405 with the nannys turned on I almost rear ended the car in front of me. As we were slowing down, the thing suddenly FLASHED a warning on the dash and my attention was drawn to that, not to the car ahead. Yikes. I KNOW HOW TO SLOW DOWN... ever since, the first thing I do when I get into a Honda car is to turn OFF the nannies.



Don't forget that it has been proven time and again that automation breeds complacency. Even if you put driver #1 in a situation where they are using automation, they are going to be more complacent on average. Just basic human nature.


 
Thread Page - 1 2 3 4 [5]
Go to:
Contact TOV | Submit Your Article | Submit Your Link | Advertise | TOV Shop | Events | Our Sponsors | TOV Archives
Copyright © 2018 Velocitech Inc. All information contained herein remains the property of Velocitech Inc.
The Temple of VTEC is not affiliated with American Honda Motor Co., Inc. TOV Policies and Guidelines - Credits - Privacy Policy
30 mobile: 0