2019 Pedestrian anti-collision systems in cars 7

[GregLocock]

Rather a tale of woe.

https://www.aaa.com/AAA/common/aar/files/Research-Report-Pedestrian-Detection.pdf

One could argue that reducing pedestrian deaths by ~10% is good, but at the same time, if the systems were working properly that'd be 100% fewer. The graph on p25 says it all, about the simplest case at the lower speed. Two of the systems at least tried to do what you'd hope they would set out to do, the other two might as well have been switched off.



Cheers

Greg Locock


New here? Try reading these, they might help FAQ731-376

http://eng-tips.com/market.cfm?

 

[IRstuff]

Fundamentally, the problem is intractable; the car stopping distance at 20 mph is on the order of 63 ft, with a stopping time on the order 2.8 seconds, but in 2.8 seconds, a pedestrian going 3 mph will have already crossed the entire width of the traffic lane, barring them actually stopping in the middle. Any closer, the car won't stop in time.


TTFN (ta ta for now)
I can do absolutely anything. I'm an expert!

https://www.youtube.com/watch?v=BKorP55Aqvg

faq731-376 forum1529 Entire Forum list

http://www.eng-tips.com/forumlist.cfm

 

[GregLocock]

That seems rather a long stopping distance. I suppose there are customer acceptance reasons why you wouldn't do a full ABS stop, whatever the morality of that is.

Cheers

Greg Locock


New here? Try reading these, they might help FAQ731-376

http://eng-tips.com/market.cfm?

 

[VEBill]

"100%"

That'll require high-powered X-rays, to detect pedestrians that are about to dart out from behind something opaque to all other sensor technologies. ;-)

Point being, existing electronic sensors (cameras, radar, lidar) are not omniscient. In the autonomous vehicle debates, too many are making precisely this false assumption that the systems are somehow omniscient.

To help clarify thoughts, if we allow the proposal that such systems can someday soon be 100% perfect, then it follows that speed limits can be eliminated. Which would be fun.

Area for improvement: As a human driver, I'm able to look under most parked vehicles watching for moving feet. In 2019, as far as I know, no electronic system includes this technique.

"...stopping distance at 20 mph is on the order of 63 ft..."

That must * include a generous dollop of human reaction time, so arguably not entirely applicable to automated systems.

* Since stopping distance is generally considered to be proportional to speed squared, if 20 mph is 63 feet, then 60 mph would be 567 feet (excessive). Thus the 63 feet must include generous reaction time.

 

[drawoh]

IRstuff,

If the car is doing 20mph (32kph) and it decelerates at 0.8g, it takes 1.1s and covers 5.1m (17ft). This does not account for the driver's reaction time.

--
JHG

 

[itsmoked]

You have to include the 2 seconds it takes before the driver looks up from their phone.

Keith Cress
kcress -

http://www.flaminsystems.com

 

[GregLocock]

The detection problem can be improved by using people's mobile phones as beacons. It is not a 100% solution, but it is a big step.

Cheers

Greg Locock


New here? Try reading these, they might help FAQ731-376

http://eng-tips.com/market.cfm?

 

[IRstuff]

There was ~2 seconds of reaction and application time included in the total stopping distance. A fully automatic system could shave over a second off that number, which would save about 39 ft.

TTFN (ta ta for now)
I can do absolutely anything. I'm an expert!

https://www.youtube.com/watch?v=BKorP55Aqvg

faq731-376 forum1529 Entire Forum list

http://www.eng-tips.com/forumlist.cfm

 

[waross]

Average Dry Stopping Distance (Assuming two-thirds of a second reaction time)
Speed 20 mph
Thinking Distance 20 feet
Braking Distance 20 feet
Total Stopping Distance 40 feet

I suppose there are customer acceptance reasons why you wouldn't do a full ABS stop, whatever the morality of that is.

I don't quite understand you here.
Admittedly my vehicles are older but we get a lot of snow and ice on the roads here.
In icy conditions an ABS stop in my vehicles takes about three times the distance of a normal stop without ABS.
Admittedly the ABS prevents a skid when the wheels lock up and it gives a second chance to back off the brakes and brake properly.
Triggering ABS and keeping it engaged won't get me stopped before the stop sign and the cross traffic.

Bill
--------------------
"Why not the best?"
Jimmy Carter

 

[waross]

63 feet may be accurate.
3 feet reaction time.
3 x 20 feet with ABS.

Bill
--------------------
"Why not the best?"
Jimmy Carter

 

[VEBill]

"...~2 seconds of reaction and application time..."

Of course it varies, but 2s is not typical.

You should remember the technique for counting off seconds: "A thousand and one, a thousand and two..."

Typical reaction time (within the normal distribution) is usually quoted as 0.8s. It's perfectly fair to round that to one second.

Two seconds is widely recommended as a 'Safe Following Distance' (measured in seconds) for good conditions. Obviously it wouldn't be very safe if it was the same as the typical reaction time.

---

All that said, perhaps two seconds is the average, with most drivers normally distributed around 0.8s, but the distracted or sleeping outliers wildly distorting the average. But that's not actually meaningful.

Outliers could distort the average for automatic systems. Just leave the clock running when they have a failure, and stop the clock (days later) after they've come back with a SW fix that passes. Average might end up as 3 hours. Not meaningful.

Point being, never embed outliers in averages without commentary.

---

In any case, human reaction time is not applicable to automated systems.

 

[VEBill]

"In icy conditions an ABS..."

I live in the forest on a gravel road that regularly becomes a sheet of ice in winter. Studded tires and 4Matic AWD are helpful.

Creeping down the icy hill, when the smooth ice is covered with fluffy snow (that can pack or 'raft' under the studded tires), is a bit scary.

In these conditions (presumably slow speed), I've noticed that my car (Mercedes) allows me to override the ABS function.

When the fluffy snow rafts under the studded tires, preventing the studs from biting into the ice, the ABS function is unhelpful. Each ABS pulse refreshes the raft of snow.

But I discovered that if I press the brake pedal slightly harder, the ABS function stops. This allows me to lock-up the wheels, which almost instantly grinds off the rafts of snow, and allows the tire studs to bite into the ice.

This response is repeatable, so I presume that this is yet another hidden and otherwise unmentioned feature.

I have no inkling of the permitted speed range of this override.

 

[waross]

When I took my first drivers test many years ago reaction time was tested. 0.55 seconds was the pass mark.
That was physical reaction time. You knew the light was going to go red and were waiting to slam your foot on the pedal.
Add some time to realize that you are facing a threat before the physical reaction starts.
ABS is better than locking the brakes and going for the ride.
Much better is backing off the brakes and re-applying a little less brake so that the ABS does not engage.
After years of driving icy roads before ABS was a feature I seldom engage ABS.

Bill
--------------------
"Why not the best?"
Jimmy Carter

 

[IRstuff]

Recommended safe following distance was 3 seconds, the last I heard anyone talk about it.

The cited article quotes a nominal 1 to 2 seconds from the stimulus to actually getting the foot onto the brake pedal and another 0.5 seconds from then to actually getting the brake engaged. This is consistent with the stopping distance calculations I've seen elsewhere. Note that physiological reaction time is on the order of 0.25 seconds, from stimulus to response, but that's in a test situation where the subjects are expecting something that they have to react to. When that imperative is removed, or substantially delayed, the reaction time increases. Moreover, most reaction time tests using hand-held buttons, which is different than removing one's foot from one pedal and placing it on another.

https://nacto.org/docs/usdg/vehicle_stopping_distance_and_time_upenn.pdf

https://law.lis.virginia.gov/vacode/46.2-880/

http://onlinepubs.trb.org/onlinepubs/nchrp/nchrp_rpt_400.pdf

A driver, using left-footed braking, could probably shave 0.75 seconds off that.

TTFN (ta ta for now)
I can do absolutely anything. I'm an expert!

https://www.youtube.com/watch?v=BKorP55Aqvg

faq731-376 forum1529 Entire Forum list

http://www.eng-tips.com/forumlist.cfm

 

[VEBill]

Google: Two-Second Rule

Then Google: Three-Second Rule

If your Google looks like mine, the latter leads to basketball.

Acknowledged that some areas recommend a three-second rule instead of a two-second rule to give an additional buffer.

Let's just agree that YMMV.

In any case, human reaction time is not applicable to automated systems.

 

[zeusfaber]

There's more to reaction time than mere "hear the bell and feel the dribble"

I won't easily forget a line from an Air Incident Report that crossed my desk some decades ago (the instrument panel had flopped down onto the pilot's lap at Rotate): "Following a delay measured in milliseconds during which I fought to suspend my disbelief...."

A.

 

[LionelHutz]

VEIBLL - There is no ABS over-ride. The ABS probably has a low speed threshold so you are likely travelling just above that threshold and the heavy braking brings you back below where it shuts off. Also, the ABS won't know the car is still moving if all 4 wheels are stopped.

 

[VEBill]

Creeping down an icy hill covered with fluffy snow all happens between 0 to 5 kmh.

The ABS would still chatter away, assisting with steering away from the edge, but allowing the speed to escalate.

When the speeds gets too high, pushing the brake pedal more firmly would reliably lock-up the wheels, scrubbing off speed at the cost of no steering.

The technique is to modulate between these two, for the minute it takes to get down the hill.

Opportunities to experience this are thankfully rare, but I've used this technique several times.

Vehicle is 2008 Mercedes E350 4Matic Sport (W211 chassis).

Excuse the thread drift.

 

[Spartan5]

I think that the reaction time in these estimates is very conservative. It's not uncommon for me to bump things from a cupboard and catch it before it hits the countertop. At less than a one meter drop, I'm recognizing what is happening, reacting, moving my hand about the same distance as my foot would have to go to get to the brake pedal, and catching the cup or whatever in less than half a second. A lot of that is muscle memory and bypasses the brain. But that applies to driving as well.

See also:

https://www.exploratorium.edu/hockey/saves2.html

https://www.nbclearn.com/science-of-nhl-hockey/cuecard/56600

 

[waross]

I have found that backing off the brakes and then applying as much brake as I dare without enough to trigger the ABS is more effective.
If you can try it safely in a few months please share your results. I will be trying the ABS on snow and Ice this winter with my wife's 2018 car. Please don't tell her of my intentions. grin
Bill
--------------------
"Why not the best?"
Jimmy Carter