Safe Speed Forums

"There are two types of Pedestrians; The Quick And The Dead "

_________________
RoADA Member -GOLD 2008
If you don't know where you are going, any road will get you there.

I did a crude "sensitivity analysis" based on the current spreadsheet.

I set these "standard parameters":

Reaction time 0.75 secs
Distance to Kerb 5ft
Pedestrian speed 5mph
Initial vehicle speed 30mph
vehicle width 6ft
braking effort 0.9g

I noted that the model predicts 6.21 deaths

Then I varied each of the important parameters in turn to produce 5.21 deaths and 7.21 deaths.

The following alterations in input parameter were required to produce the 2 deaths range:

DRT: 16%
DTK: 15%
PS: 22%
IVS 7%

This appears to show that the model is more sensitive to initial vehicle speed than any other parameter. I don't actually believe that this is very realistic, but I can't easily prove the case at present.

One problem is that the proportion of deaths is far higher than anything that takes place in the real world, and we don't really know why.

Another might be that a 7% variation in speed is pretty significant while at 16% variation in reaction time is nothing - real world reaction times probably vary over a much wider range. maybe from 0.4 seconds to 2.5 seconds.

_________________
Paul Smith
Our scrap speed cameras petition got over 28,000 sigs
The Safe Speed campaign demands a return to intelligent road safety

I've updated the spreadsheet to include a deaths calculation based on the "Minnesota" curves for adult pedestrians.

Pedestrian Incursion spreadsheet:
http://www.safespeed.org.uk/pedinc.xls

Minnesota report:
http://www.lrrb.org/pdf/200223.pdf
(The chart in question is Figure 2.2 on PDF page 32, page number 20)

With "standard parameters" the Minnesota method suggests 6.6% of pedestrians hit will die.

Official DfT figures for pedestrian injuries and deaths on built up roads (30mph and 40mph speed limits) are as follows:

killed: 572
all injuries: 34,981
killed percentage: 1.6%

from: RCGB Table 24

What the DfT figures fail to show are:

* pedestrian deaths caused by reckless behaviours (reduces the 572 figure when we're talking about normal motorists)
* pedestrians who were hit, but uninjured and unreported (increases the 34,981 figure)
* pedestrians who were hit and injured but unreported as a road crash (increases the 34,981 figure)

If we make some guesses for these values (-72, +30,000, +30,000) we get:

Killed in crashes with normal motorists: 500
involved in incidents: 94,981
killed percentage: 0.53%

Even using the Minnesota curves (4 times more optimistic than the A&M curves) the model is still out by a factor of 12. (6.6%/0.53%)

And we haven't even looked at near misses where emergency action saved the day.

Clearly something is terribly wrong...

_________________
Paul Smith
Our scrap speed cameras petition got over 28,000 sigs
The Safe Speed campaign demands a return to intelligent road safety

I found a bug in my use of Excel's ill documented "PERCENTILE" function. I've corrected it and added a page to the spreadsheet that shows the A&M and Minnesota curves.

The effect of the bug was to exaggerate the risk of death. Now it's fixed the results are closer to reality - but nowhere near close enough for my liking. The Minnesota prediction with standard parameters is now 3.3% deaths, whereas my best estimate of reality is rather less than 0.53% - it's still out by over 6 fold, and I've done nothing to exclude 40mph speed limits!

_________________
Paul Smith
Our scrap speed cameras petition got over 28,000 sigs
The Safe Speed campaign demands a return to intelligent road safety

Your model implicitly assumes zero pedestrian reaction. In reality, a proportion (most?) pedestrians who set off on a path which results in an impact will not be totally oblivious to the fact they're about to be hit but will take some avoiding/mitigating action to reduce the severity of the impact. Thus (I assume) the official figures will include pedestrians who are hit by a vehicle but the contact is a glancing blow or perhaps just contact with a door mirror?

Correct. On the other hand we have zillions of near misses that aren't predicted by the model either.



I played with that by reducing the width of the vehicle. If we assume that overlaps of 6 or 8 inches represent no collision at all, we still get funny results in terms of killed percentage.

_________________
Paul Smith
Our scrap speed cameras petition got over 28,000 sigs
The Safe Speed campaign demands a return to intelligent road safety

As you have identified a discrepancy in the predicted deaths results, would it not be better to test the sensitivity of "actual impacts"?

That sounds like a great idea, but it doesn't work (I tried it and found a limitation). It's to do with the "car width". If we have slower reactions for example, the "impact width" remains the same.

I'm now playing around with the area under the impact graph.

Thanks for triggering that off. I'd done it before, but I'd never looked at the sensitivities.

_________________
Paul Smith
Our scrap speed cameras petition got over 28,000 sigs
The Safe Speed campaign demands a return to intelligent road safety

Suppose you incorporate an assumption that a fatal impact only occurs where a pedestrian is located in the middle (say) one third of the car's width. In other cases, the pedestrian will have either leapt back or leapt forward (only to be struck by a car coming the other way?,) or the car has swerved, sufficiently to mitigate fatal impact.

It's certainly worth playing with.

I'm not sure that we can make clear enough assumptions to learn anything much from it. It's going to be very arbitrary isn't it?

_________________
Paul Smith
Our scrap speed cameras petition got over 28,000 sigs
The Safe Speed campaign demands a return to intelligent road safety