Sunday, 30 August 2015

Bristol vs Self driving cars

One little sighting over by our strategic PCSO-state partners, Google, are their self driving cars.

They are utterly not-ready for Bristol. You can see this at a glance

Not only is there a bollard on the roof, there appear to be sensors coming out of every corner.

Here in the city centre we have a word to describe vehicles with wing mirrors attached: visitors.

While having wingmirrors helps you in certain out of town operations, primarily changing lanes on the M4, in town it actually hurts you: it makes your car about 30 cm wider so significantly reduces your choices as to where to drive. And, as their presence broadcasts that you are not a local —but instead have a vehicle you care for— you lose every negotiation that takes place, be it a junction or a "who will give way first" interaction. And, when parked, you don't have to bother looking out the window when you hear the sound of a car-on-parked-car interaction, unless it is so loud that you fear it may be some damage needing bodywork repairs.

Now put something on every corner of the vehicle whose presence is actually critical for the self-driving feature to work. Sensors attempting to detect what is too close, so the multi-layer neural network that is is google self-drive program can make a decision as to what to do next.

It's doomed.

That's before even looking at what complexity of road the cars are exposed to. This road, "Castro", is a complex environment on the basis that it actually has people walking. Yet it is wide-enough for oncoming vehicles to pass each other, all junctions are nice simple right-angled crossings, and visibility is reasonable.  Now imagine taking a car trained here and trying to drive down Clifton Vale, the one with the blind z-bend you have to share with oncoming traffic. The car would just give up. And it would share the experience with every other google car, saying "avoid this road". Before long whole swathes of the city would be blacklisted by Google cars, cars who give up on account of their paintwork and external mounts being valued.

Which, when you think about it, could be no bad thing

Saturday, 29 August 2015

where are the speedo-watching crashes?

Here are the two petitions to look at

Speeders: Scrap the 20mph limit in Bristol and restore common sense  now at 8045
Keepers: Keep and extend 20mph limits now at 1680

One claim in the Speeders' petition is
"roads will only be made more dangerous with frustrated drivers and people watching the speedo rather than where they're going!"
We haven't noticed that ourselves, as we have a special speed limiter in our car called "the gear stick". Put it into "speed band 3", or "3rd gear" as it is sometimes known and it burbles along quite happily at 20-24mph, except in the special case that someone drives so close that you can see the driver's nostril hair and you slowly lift your foot of the accelerator to drop to precisely 19 mph just to see the expression on the driver's face in your rear view mirror.

Anyway, it may be that the speed petitioners don't have manual transmissions, or are somehow unable to determine road speed, and do have to look at the speedo all the time.

This is something we can test now that the central 20 mph zone is over a year old. All we have to do is look at the data.

One irony here is that if they do occur, the fact that they will take place at 20 mph means the police aren't likely to get involved. Any crash involving pedestrian, cyclist or stationary object will be less likely than at 30 mph to injure anyone, so they may not get called out. Indeed, even the insurance companies may not get a look in if people care about their no claims discount. head-on collisions will still have 40-50 mph of energy, so are more likely to show up. This means official data sources: —police and insurance— may show a reduction in RTCs even if there has been an increase in RTC events.

We ask the speeders, then: where are the speedo-watching crashes?
  1. How many crashes in 20 mph zone have you personally been involved in where you or another participant was looking at the speedo at the time?
  2. Have you heard of any such crashes —and do you have the contact details for us to follow up on this?
  3. Is there any other evidence for an speedo-watching crashes in the central bristol zone?
We've had the zone for 18 months now, with millions of journeys in it by now. If speedo-watching crashes are a risk in 20 mph zones, we should have seen some. 

If we don't have the any signs of speedo-watching crashes, then, it could be due to
  1. They are happening, but the massive lower energies in the collisions cause them to be unreported
  2. That whole claim about "watching speedos causes crashes" is bollocks.
#2 is the null hypothesis: 20 mph limits do not cause speedo-watching crashes, is the one which has to be disproved to a significant degree of statistical confidence before it can be believed.

We put it to the speeders then: show us the data.

Now, assuming, on the off-chance that there is the data, that speedo-watching does cause crashes, hence 20 mph zones are more hazardous than 30 mph zones, why do the petitioners still propose lower limits by schools?

Either 20 mph zones are more dangerous due to speedo-watching or 20 mph zones are safer round schools.

So if that claim "20 mph is more dangerous" isn't utter bollocks, then, if the speed campaigners really believed it, they should be pushing for 30 mph zones round schools "for the children".

Speeders: show us the data —you've had 18 months to collect it.

(photo: kid scootering to school outside Christchurch School, Clifton, pre-RPZ)

Friday, 28 August 2015

Unhappy speeders

In our ground breaking analysis of the geographic distribution of the speeders and the 20 milers in the city,  we were picked up on for our statement "They must live very unhappy lives."

Good catch. Judging by the reaction to the post, we should have concluded "they are very angry people"

It's as if they spent a lot of time in traffic jams or stuck at the lights -blaming George Ferguson for every minute of their wasted life

We were particularly called out on our assertion that the 38% of out of town petitioners on the "right to speed" campaign don't count.

Specifically, the accusation of misinformation came about because of the T&Cs of the council's petition policy, which states
If your petition has received 3500 signatories or more from people who live, work or study in Bristol it can then trigger a full council debate [see page 5] and if this is the case we will discuss with the lead petitioner the options for enabling this to take place.
We are not attempting to misinform anyone. Look at what we wrote
Of the speeders, 38% of them don't live in in Bristol. Which means they are, as far as Bristol elections are concerned, as relevant as residents of the Isle of Wight. They don't have a vote, all they have is a whine.
See that? If you live outside the city, you are electorally irrelevant. Which may or may not transfer into the decisions about the region and its transport policy.

If you lived out of the city, you wouldn't have got a bit of a paper asking if you wanted a referendum on having a mayor —you didn't get a say. You wouldn't have got a bit of paper saying "who do you want to be mayor" —again: you don't get a say.

The fact that the council has a policy for petitions is something to cherish. The fact that they even let people from outside the city add their names shows that we do value those people who live out of town. But when it comes down to whom the council has to prioritise, it's the residents who vote for the councillors and mayors.

Everyone outside gets to make a whining sound, either in their own home, the BEP web site or their car sitting on the M32.

Is that fair? Maybe. Is it functional? Not for a region wide transport policy. But here's the problem: the N Somerset and S Gloucs councillors like their little kingdoms too much to share them.

Here we see Elf-King App Rees switching from demanding that the Clifton RPZ be removed screaming that George F is trying to dictate parking policy in Leigh Woods.

He loves being a small fish in a very small pond, and any attempt at having a broader region for  democratic governance as a threat.

Is S Gloucs any better? Well, they are very proud of the the fact that they are not quite Bristol, even to the extent of having a "Welcome to South Gloucestershire" partway along the Filton Road weekday traffic jam. Because Filton is, after all, distinct from its neighbours. But they do ask staff at the N Fringe of the city for their input on the latest bit of random roadworks.

What is not clear, though, is Why is Filton out of Bristol? . Same for those bits in N.E. Bristol. Emerson's Green, Rodway Common, etc. Part of the featureless hinterland of the city. And yet: you don't get a say —only the right to get angry about things happening in a city nearby.

Those pro-speed petitioners: do their opinions count? Not for 38% of them, no.

We've stated repeatedly we are Bristol's premier data-driven transport new outlet, compared with the evening post, which is driven by "what gets the most paper sales to our dying customer base" and "what generates the most page hits". Controversies involving parking, cyclists and speed zones hit all three.

Sadly, we don't have access to the BEP customer dataset or the details on commenters they extract from their linked-up google accounts. What we do have, however, is the python code needed to convert the published signatory list into a CSV file, with some extra flags to indicate whether or not the petitioner is in a 20 mph zone or not.

            20 mph ward  rest of BRS  CUBA    other 
Pro            920         305         116     201
Speeders      1552        3365        2289     705

As a graph, showing the numbers by area, things become more obvious

Of the speeders, 38% of them don't live in in Bristol. Which means they are, as far as Bristol elections are concerned, as relevant as residents of the Isle of Wight. They don't have a vote, all they have is a whine.

Looking at Bristol itself, we see a marked split between those people in wards with 20 MPH zones vs those which aren't.

Even though the pro-20 MPH petition is a fraction of the size of the speeder's one, it is not far off having 40% of the total petitioner count from the 20 MPH zones themselves.

This implies some things
  1. The people who get most worked about 20 MPH zones don't appear to live in them.
  2. Many of the people who get worked up about Bristol's 20 MPH zones don't even live in the city.
  3. They must live very unhappy lives.
  4. A lot of the people in the 20 zones seem pretty happy with the zones and their lives.
We'll collect some more data next week, and make up some new conclusions. Until then,  get out there and get some signatures for whichever petition you care about.

Monday, 24 August 2015

The 20 mph war on our council web site

There's now a battle of petitions up on the council web site

The scrappers: , "scrap the 20 mph limit and restore common sense",

We love the use of "common sense" in the title. "Common sense" means "obvious to the person making the statement". Yet, as we look throughout history, "common sense" meant the sun went round the earth. Galileo's Heliocentricity hypothesis was considered so heretical the common-sense regime at the time (Catholic Church) sentenced him to death. That's what common sense means: superstition over rational thought.

The retainers, "Keep and Extend 20mph limits"

These want the limits retained, extended, and maybe even enforced. We will call these "the people who believe Newton's equations about momentum and kinetic energy".

There we have it then, two factions: those who reject the physics of RTCs on the grounds of "common sense", and those who care about people walking and cycling round the city.

It will be interesting to watch the numbers. the flat-earthers have the support of the local paper, and a three month head start in the petition.

As of June August 23:

Flat earthers: 7421
Progressives: 1053

This is something we'll be keeping an eye on.

Saturday, 8 August 2015

Traffic Cameras: they don't shoot black americans

One of the recurrent themes in the UK press, with the Daily Mail-esque "Speed Cameras steal from motorists" story also recurring in the US,

Usually there are people quoted demanding instead real police, in real police cars, catching people "really" breaking traffic laws —by which presumably they mean "people other than themselves"

What doesn't get covered, though is how scary it is to be stopped by the US police compared to those in the UK. In the UK you get pulled over with some opening gambit like "is this your car, sir?", or "how much have you had to drink this evening?".

In contrast, in the US the lights go on so bright, with such a loud siren that you look up from your phone with a jerk, to see the police car filling your mirror and pull over in fright. Only here, the police officer (almost invariably a white man) walks over with their hand near their gun, ask you to wind down the window and then keep your hands visible, all the while standing a bit behind the door so that you are in front of them and its easier for them to shoot you than vice versa.

It is frightening, and you have to be so careful not do to anything not to make them overreact, and reach for the gun. And that's just as a white adult male in a nice car in a nice part of the US.

Now imagine you are an 18 year old black american being stopped at night. You know the police have a track record of shooting black americans even when pulled over for "minor traffic misdemeanours".  This is a country where failing to indicate as you change lanes has a death penalty.

You are going to be scared whenever you get pulled over, if you look at the statistics from places like Ferguson, you will be pulled over a lot more than white americans, and the is a small but tangible risk of you not getting out the stop alive.

In contrast
  1. Traffic cameras don't pick on drivers because of their skin colour or age/state of vehicle.
  2. Traffic cameras don't shoot black americans. Or any american, to come to it.

While in suburban americans may scream about unjust and unfair traffic cameras, you might find some locals with different opinions. Ask black americans kids whether they'd prefer being stopped by the police with a risk of being shot or having traffic violation fines sent in the post —you may find more support for the cameras. Now ask their parent, those parents who worry about their children going out for drives at night in the US, and they may give a very different answer.

Traffic Cameras: they don't shoot anyone

Visiting our strategic partners in Mountain View, California

We've stated before: with our strategic partners Google and Facebook we are building a PCSO state, one where people pay month for the privilege of being monitored, such as the how your Android phone reports in your movements 7x24.

We do of course have to visit our partners sporadically to brief them on developments, hence a short trip to the US. Expect some commentary as well as insights from Silicon Valley, which is now the hub of the future automobile. Even watching people gives us a profound vision of the futre

First, how to carry furniture in a "convertible"

The correct approach, clearly is to stick the chair in upside down, with someone on the passenger seat to keep an eye on it. The driver themselves gets some information on the chairs status: if they look in the wing mirror and see daylight then the chair has gone. If they hear over their music the sound of brakes and crash, they may also get an audible cue of the "chair loss event"

Until then, luggage transportation at work.

We just have one outstanding question though. We've seen them carrying a chair. What are they going to do with the Sofa?

Tuesday, 14 July 2015

Cyclists please dimount

We don't know what dimounting is, but given the incident near Bristol where a woman crashed her car while playing with an entertainment device, we have our suspicions

This is Temple Back East from Temple Back bridge heading for Temple Meads station from Old Market roundabout.

It could be a mis-spelled "cyclists dismount" sign, except there's no reason to dismount, not when there's the hatching to the left.

Dimounting, then, must mean something else. Our theory: riding a saddle designed to entertain your entertainable bits while on a ride.

Wednesday, 1 July 2015

Daylight running lights: the helmets of cars

It's been mandatory since 2011 for cars to have daylight running lights ("DRLs"). Is this a good thing? Or just it just up the arms race between motorcyclists, cars, pedestrians and cyclists?

For a change, there's actually some research on this topic, in the form of a 2008 study from the US government, "The Effectiveness of Daytime Running Lights For Passenger Vehicles". Bear in mind that the sample set of this study is of course US drivers, who are a lot more common than US ones to drive round at night without lights on. DRLs partially mitigate for this inattentiveness —or it may amplify it if drivers don't notice they have their lights off, on account of the DRLs illuminating in front of them.

The analysis evaluates the effects of daytime running lights (DRLs) against three types of target crashes: (1) two-passenger- vehicle crashes excluding rear-end crashes, (2) single-passenger-vehicle to pedestrians/cyclists crashes, and (3) single- passenger-vehicle to motorcycle crashes. Each crash type was examined at three crash severity levels – fatal, injury, and all severity. The basic approach is a control-comparison analysis of real-world crash involvements for DRL-equipped vehicles and non-DRL vehicles. Ratio of odds ratios were used to derive the DRL effects. A 95-percent confidence interval was used to infer statistically significant conclusions. The Fatality Analysis Reporting System and the State Data System were the crash data sources used for this analysis.
The analysis found that DRLs have no statistically significant overall effects on the three target crashes. When combining these three target crashes into one target crash, the DRL effects were also not statistically significant. When examined separately for passenger cars and light trucks/vans (LTVs), DRLs in LTVs significantly reduced LTVs’ involvements in the target two-vehicle crashes by 5.7 percent. However, the remaining DRL effects on these three target crashes were not statistically significant. Although not statistically significant, DRLs might have unintended consequences for pedestrians and motorcyclists. Particularly, the estimated negative effects for LTVs were relatively large and cannot be completely ignored.
The final two sentences are the highlight: from the survey done 8 years ago, it appears that daylight running lights may make it more dangerous for pedestrians and cyclists. And what are LTVs? That's pickups and SUVs.

The study was conducted by looking at crash statistics and seeing if those vehicles with DRLs were in crashes more or less. It wasn't a randomized sample (give people cars with DRLs on or off), which could skew the numbers (drivers of volvos are safer than drivers of ford mustangs, etc, older cars have worse brakes, etc.). Furthermore, as whether full lights were on or off wasn't recorded, the impact of DRLs at dawn or dusk can't be established.

Going in to the details, skip the maths unless you want to learn more about statistics., go to P4.10

based on the combined PC (Passenger Car) and LTV (Pickups and SUVs) results, DRLs seemed to have no overall effect on daytime target crashes.
See that? No statistically significant affects whatsoever. They are the helmets of cars: useless, yet designed to make you feel bad for not having them.

Now, P 4-14
For all crashes, again the estimates were not statistically significant. However, based on the combined State data, DRL effects for both PCs and LTVs were negative. Overall, DRLs seemed to increase Single-PV-to-PED/CYC crashes by 5.6 percent.

There you have it then. A feature which does nothing for overall safety rates, but which actually appears to increase risk for pedestrians and cyclists.

And on P 4-16, motorbikes
DRLs seemed to increase daytime Single-PV- to-Motorcycle crashes by 1.2 and 17.3 percent for PCs and LTVs, respectively. Overall, DRLs seemed to increase daytime Single-PV-to-Motorcycle crashes by 5.0 percent. None of these effects were statistically significant.
There you go then: daylight running lights appear to slightly increase the risk to vulnerable road users, while delivering no benefit whatsoever for the occupants of the car themselves.

Given this data: they do nothing for cars and endanger others, why offer it, and then why mandate it?

Offering it in cars could be assigned to a belief that it does make a difference (possibly even some unpublished data that it does).  Or, more cynically, because if possible purchasers ares somehow convinced that DRLs improve their safety, then they would buy that particular model of cars, possibly paying a premium. If you are a car manufacturer who discovered that spending £1 to ensure your car lights stayed on all the time could earn you £100 —you'd do it.

But why mandate it? That would remove all differentiation between vendors?

Some hypotheses(*)

  1. Politicians also mistakenly believe there is a safety benefit, and that by mandating it they can improve safety.
  2. Politicians don't care where it is beneficial or not, they just want to be seen to be doing something that they can take credit for and doesn't cost much.
  3. It is used as an excuse to avoid fundamental changes in road safety. Here the car manufacturers would be lobbying politicians. Unlike other features (ABS, Airbags) which actually cost money, DRLs are cheap to roll out. You'd push for them if the alternatives cost money.
Whatever the reason, its motorcyclists, pedestrians and cyclists who appear to lose out.

(*) There is also the theory that there is some other work that through some controlled A/B test shows that it does make a difference, and furthermore politicians have read and understood the details. This hypothesis is considered significantly unlikely.