Tag: cycling

Blindspin 3: Why would someone want to volunteer?

 

Volunteering to be a test subject for Blindspin (project page) means that you may spend an hour or two of your life bicycling back and forth on a track and looking weird, probably picking up a heavy sweat, and potentially risking your personal safety. Why would anyone do that?

Short answer: because no one else has ever done this. Science means that you methodically pile bricks to create something useful. An individual brick is just a turd-colored clump of burned clay, but they all need to be in place. And in this particular niche of cycling safety, no one has really even picked up the first block.

What is the niche, and why is it interesting? The idea for this research spun off from work we’ve done in driver distraction. A large number of accidents are caused by the driver’s attention wandering to something irrelevant (these days, largely using a mobile phone). Yes, there are brute-force political ways to handle that particular problem, such as outright bans on mobile phone use (which don’t really work very well).

But if we actually want to approach the problems scientifically, we must ask simple-sounding questions that are measurable. In this particular case: how do we measure what driver distraction even means? How do we find numbers that allow us to compare how dangerous driver distraction would be in different scenarios?

We need those numbers before we can even think about how to think about answering more directly practical questions, such as: how do we design car interfaces so that they do not cause dangerous distraction; how do we intelligently attack the problem of driver mobile phone use; and even, how do we design roads (especially intersections)  so that if and when distraction does occur, the impact will be as small as possible?

For cars, a lot of research has been done, and some of it has found its way into safety recommendations (very slowly, but steadily). With cars, there are well-established safeguards that can alleviate the effects of such distraction — lane markings, brakelights, and standards for traffic light design, for example. For cycling, such safeguards have been far less systematically studied.

staubin-design

The science of car road design. (Source: paulstaubin.ca)

bridgewater3

The reality of cycleway design (Image source: madcyclelanesofmanchester)

Continue reading Blindspin 3: Why would someone want to volunteer?

Blindspin 2: How to do science by dumpster-diving

When a project has a zero budget, everything has to be hacked and improvised, typically with duct tape. The data collection system for project Blindspin is a good example. (For a project description, see “Does it make sense to ride a bike with your eyes shut”)

If we had a budget, we would be looking for a high-end mobile data logger with millisecond accuracy. Since we don’t, we would very much prefer to use smartphone that someone has thrown away. And it seems we can.

The basic requirement is not very complex. Our system will consist of a pair of electronic goggles which are normally opaque, but which the cyclist can turn transparent by pressing a switch. We need to record the time of the keypress, hopefully with millisecond accuracy. We also need to collect GPS location information, so that we can determine the path the cyclist drove while blinded.

There are GPS data logger apps galore for Android. We found that the AndroSensor software is almost perfect. It collects GPS location, accelerometer information, and all other sensor data at resolutions of up to 50 milliseconds. The only problem is how to input information about the key presses. There is no sensor channel for that.

However, we realized that AndroSensor can record the ambient sound level in dB. So we decided to use the audio channel to store button data. In the simplest case, button down (vision occluded) is a loud noise, button up is a quiet noise.

A major problem is that AndroSensor (and most other software we looked at) always uses the phone’s own microphone, even when a line in is used. Thus, it is necessary to input the noise directly into the external microphone.

For a pre-test, we came up with a somewhat rubegoldbergish approach, but one that works. To generate the noise, we used an aviation scanner that has a reasonably large tangent button. The scanner’s autogain means that if there are no aviation transmissions, there is no sound output. However, if the tangent button is pressed, the noise is heard. The gain can be set so that the difference in noise levels is tens of dB.

To eliminate outside noise, the speaker was attached to the phone’s microphone with  Blu-Tack (sinitarra), and the whole thing covered with more Blu-Tack. Thus, the microphone hears almost no external sounds at all. When the tangent is pressed, it hears the noise from the scanner, at tens of decibels.
Blog_medium
The speaker is buried within the mass of Blu-Tack and pressed directly onto the phone’s microphone.

The whole system was attached to the bike handlebars with zip ties. Several different setups were used; the one in the picture below is operated by the index finger. A simpler way was to mount the scanner facing the other way and below the handlebar, so that it could be operated by the thumb.

 

Blog_Bikepic

 

The ergonomics of the system are horrible. However, at this stage it is simply needed to demonstrate the data collection method. Subject A tested it on a straight road and a curved one. Whenever A closed his eyes, he pressed down on the tangent. Whenever he opened them, he let go of the tangent.

The image below is the first data ever produced in this project. The red blue line is the speed given by AndroSensor. The red lines are the dB levels.When the red line is above 60 dB, the eyes were closed and the tangent was pressed, and thus the scanner outputted noise directly into the phone’s microphone.
Plot1

 

We have collected more data from subject A, but will not release data yet. Why? It is not ready yet, but even more importantly we don’t want to skew the results that the volunteers might get. Such ignorance is almost always desired in human research (with the exception of self-testing). The volunteers should have no idea whether it is possible to keep the eyes shut for one second, or for twenty — and they really should not know what we are even looking for at this stage.

For subject A in this particular case, he had a total of 9 occlusions within a 20-second period, with approximately 500-ms eyes-open periods. The occlusion times are between 1 and 2 seconds for this subject for this point on this track for this setup. The occlusion times may be longer in other circumstances… or they may be shorter.

In the final application, we will use a somewhat more complex keypress arrangement, since we will be using an Arduino to control the system. Most likely, we will use a buzzer to create an audible signal about the state of the goggles (loud buzz when goggles opaque, silent when goggles transparent).

The specs of this system are not ideal, but they are actually good enough even for real science. Data are stored at 50-millisecond intervals, but even in simulators, typical intervals are 100 ms. The biggest problem is timing the key press; even if we can get a perfectly sharp rising edge, we will have a 50-millisecond uncertainty in the timing. In practice, it may be even larger if the edges are not completely sharp. We can thus reasonably expect to get a 100-ms time resolution, but not much better. That will be sufficient, as long as we are careful to note it in the analysis.

Of course, this system does have major disadvantages, such as unknown delays in the phone software. We will design a better system if at all possible. But this is a fallback solution, which in the very worst case we can use as the actual solution. Costing zero euros.

See also Blindspin project page.

 

 

BLINDSPIN 1: Does it make sense to ride a bike with your eyes shut?

We are starting a research project which sounds completely insane but is not: we want to know what happens when a person rides a bicycle with his eyes shut.

The idea is not completely insane because the same principle — the visual occlusion method — has long been used in automotive safety research.  It is a powerful tool for studying driver distraction.  Most members of our ad-hoc team have scientific experience with the technique (but we are working in our free time on this project). First, some background.

Collaborators: Jakke Mäkelä, Niko Porjo, Tuomo Kujala, John Senders

What is the point?

Bicycling has been around for more than a century, but how much do we actually know about what could make the actual process of cycling safer? At the moment, cycling safety is more or less all about passive safety: cycleway design, design of crossings, helmets, visibility, audible alarms.

What about the process of cycling? What makes one cyclist safer than another? Are the elderly risky drivers, or the very young? What kinds of situations are the most dangerous? How and why does cycleway design affect cycling safety? (For a critical look at some cycling safety issues, see post “Does it make sense to bike without helmet?“).
beseen_be_heard395x305

Passive driving safety. This is the normal approach. Source: Ontario Ministry of Transportation

 

Those questions are too wide to be answerable, and it is necessary to narrow the focus. One area to look at is driver distraction, if only because no one has really done so. For automobiles, driver distraction is a serious study subject and has its own conferences. Cyclist distraction is only now emerging. The scientific literature seems to consist of just a handful of papers from the last year or two, and almost no field tests.

We want to ask a very simple question: how dangerous is cyclist distraction?  There are lots of accident statistics around, but they tend to look at the cyclists as victims. To put it politically incorrectly, we want to know about the cases in which the cyclists are “perpetrators”.

[Of course one never actually uses that term in research. When trying to study or prevent accidents, it’s pointless to “blame” anyone. Mistakes are made, accidents happen, and the important thing is to understand the mistakes so that further accidents can be prevented. But the term “perpetrator” is easy to understand].

This needs to be narrowed further: what do we even mean by “cyclist distraction”? Does a cyclist need to keep his eyes focused on the road all the time? Is it safe for him to glance at his watch? Could he safely talk on the phone (with a handsfree set of course)? We lack data on basic questions like this.

 

phpThumb_generated_thumbnailjpg

Distracted cycling. How dangerous is this in reality?  Source: Youthforroadsafety.org

 

What is the visual occlusion method, and what can it tell us?

We will write more on this in the next few weeks. In brief: an occlusion test determines how tightly a driver needs to keep his eyes on the road. Not every glance away is equally dangerous, and short glances happen constantly (for example, glancing at the speedometer takes about half a second).

In practice, the driver wears a pair of special electronically controlled goggles which can be quickly switched between transparent and opaque modes. A few companies manufacture such goggles, but since we have no budget, we will hack ours from a pair of 3D glasses.

At any given moment, the driver can control whether he sees the road or has to drive blind. The driver has full control, and drives at a level where he feels no danger; any accidents would make the tests invalid, since the aim is to look at safe behavior and not risky behavior.

A video shows how this was done in the 1960’s: Pioneer Days on Rt 128. A very funny yet very serious article on the subject has been published in the Boston Globe. John Senders, the principal investigator in those experiments, is also a member of our team — in fact, he is the one who proposed this experiment in the first place.

 

HELMET CLOSED 1963

Occlusion tests in the 1960’s. Photo Michael Dwyer/AP. Source: Boston Globe.

 

Since those days, laboratory simulators have been used because there are some legal and ethical issues in driving blinded on normal roads. However, it is difficult to simulate bike driving in a laboratory. We will need field tests.

The results from an occlusion study can be rather obscure, and not simple to explain. Roughly, we expect to find two critical values:

*The time duration above which it is definitely unsafe to keep the eyes closed (or focused on something else). For car driving, this time is about 1.5 seconds. We have no idea what it will be for cycling.

*The distance which a cyclist is able to move with his eyes closed. This is typically 5-20 meters for car driving.

These numbers cannot be used immediately for practical applications, but they are necessary background. The time could be used to evaluate whether (for example) mobile phone use is less dangerous or more dangerous than when driving a car;  the distance could help in designing cycleways, since it gives an indication of what types of obscurations are particularly dangerous.

Who is doing this, and why?

Currently, our ad hoc group has four people with research or technical backgrounds.  It is difficult to get funding or ethical approval for a project of this type, so we are doing it without any. This is a spare-time effort without any input from our employers.

All of us work in our spare time because we believe this research is interesting and important. Because there are risks involved, we will mostly need to self-experiment. However, to gain a large enough data set, we are seeking volunteer test subjects who have moderate streak of lunacy.

*Jakke Mäkelä (LinkedIn) worked in an automotive safety research project in 2013-2014, and is familiar with occlusion methods. He is in some vague sense the unofficial project leader, to the extent that there is one.

*Niko Porjo (LinkedIn) is a technical wizard. He will be hacking the occlusion goggles and working on data collection.

*John Senders (home page)  is one of the pioneers in the field of visual demand (and is featured in the video above). He proposed the idea of studying cycling visual demand in the first place. He will work on theoretical aspects in particular.

*Tuomo Kujala (LinkedIn)  has studied visual demand in automotive environments, and has done extensive visual occlusion studies. He will try to relate this project to earlier research and will work on data interpretation.

We are open to adding new people to the core team, especially experts in cycling safety.

 

What do we plan to do next?

1. First and foremost: we plan to self-experiment and take the personal risks before we allow anyone else to participate. We don’t think there are any real risks, but… We will report our results on this blog as we get them. The figure below shows some pre-pre-pilot results, but those mainly show that the data collection method works. Critical technical parts of the experiment are still missing.

Pic1

 

Results from pre-pre-pilot, subject A. A drove along a straight isolated track of cycleway, and pressed a button whenever his eyes were shut.  The blue line is speed in kmh. The red lines are above 12 when A kept his eyes shut, below 8 when he had them open. 

2. We plan to seek volunteers in July-August 2014 in the Turku region. Self-experimentation is useful up to a point, but it does not give a large enough dataset. The experiment is actually much safer than it sounds, but it does require a certain amount of craziness in the volunteers. We already have a few such volunteers lined up, but we could almost certainly use more.

3. We plan to publish the results. Even though this is a free-time zero-budget project, we seriously aim to get peer-reviewed results. If we cannot get published in a peer-reviewed scientific journal, we will publish on this blog.

If you are a moderately insane cyclist in the Turku region, and are potentially interested in volunteering, please contact Jakke.Makela{at}gmail.com.

See also Blindspin project page.

 

 

Does it make sense to bike without a helmet?

The blog post “Why it makes sense to bike without a helmet” is giving me a headache. It is wrong, wrong, wrong, yet it’s surprisingly difficult to point out exactly why. The author argues that  “if we start looking into the research, there’s a strong argument to be made that wearing a bike helmet may actually increase your risk of injury, and increase the risk of injury of all the cyclists around you.”

The author essentially argues that by sacrificing some personal safety now, he can improve the safety of everyone in the future. That is a laudable attitude. But is he actually doing that? I am becoming more and more interested in cycling safety as I am turning greener and greener. Thus, this needs to be analyzed out.  A faulty argument in favor of a good cause is not acceptable.

The author cites an impressive number of statistics, but the arguments seem to be quite simply invalid. Correlation and causality have been confused, and so on. Multiple errors. It would be easy to shrug it off, but the post has been shared and discussed widely.

Also… it’s way too lazy to just sit on the sidelines and criticize. The author bravely went out on a limb and said something controversial, even though it seems he’s completely wrong.  So, here’s a counterquestion that respects that bravery: are there any conditions under which he would in fact be correct?

The logical chain

Here’s my reconstruction of the main logic of the blog. These are not the exact claims of the author, but something that can be inferred from the text. The mathematical additions are mine.

1. Helmets decrease the risk of serious injury, if a cyclist has an accident. This is a Bayesian variable: p(S|A). p(S|A) is smaller if one wears a helmet.  The probability of severe injury is then p(S)=p(S|A)*p(A)

2. Currently, the probability p(A) of being in an accident is relatively high when cycling. For someone who cycles a lot, it is probably in the range of 1% per year (my estimate).

3. If cities were optimized for biking, the probability of an accident p(A) would be much lower than it is now. Biking might not be any more dangerous than driving a car or walking. At that point, it would be irrelevant whether or not one wore a helmet.

4. To force cities to be optimized for biking, one must motivate the maximum number of people (N) to cycle for maximal amounts of time (T); that is, maximize the amount of cycling, C=N*T. The larger C is, the smaller p(A) will be.  For future reference, note that C can be considered to be general measure of how attractive cycling is perceived to be.

We don’t really know how to model the effect. However, for lack of a better model, we could assume that it follows the exponential distribution p(A)~f(λ,C)=λ*exp(-λ*C) which has mean 1/λ. Since we can scale the constants freely, let us set λ=1. Then, the current probability of an accident is P0=exp(-C0). We want to evaluate how the probablity changes as C changes.

5. Mandatory helmet use is likely to decrease both the number of cyclists, and the time used for casual cycling. We can call this the F-factor, as in “F you”, where F<1. Then the accident probability given mandatory helmets is p(F)=exp(-C0*F) = P0^F.

Rough estimate: if the current personal probability of an accident per year is 1%, and a mandatory helmet decreases cycling by 10% so that F=0.9, then the mandatory helmet would raise the personal probability to (0.01)^(0.9) or 1.6%.

6. Therefore, mandatory helmet use will slow down the target of creating a biking-optimized city, and increase the probability of being in an accident. Up to here, the arguments may actually be valid. However, now it starts to break down.

What is missing 1: Going from big F to little f

There is a problem here. Whether an individual wears or does not wear a helmet does not have any bearing on whether the government does or does not make helmets mandatory.

The author seems to imply that using a helmet is “giving in”: it is a signal to society that cyclists can be trampled on. This sounds vague, but let’s model it in any case. We could consider such an effect to be similar to the F-factor, in that it makes cycling less attractive to everyone. We can even model it similarly, calling it small f.

Using a helmet would thus increase the probability of being involved in an accident to P0^f. Note that by our definitions, f is larger than F; a small effect means that the value of f is close to 1.

What is missing 2: going from probability to risk

Why does this sound completely unsatisfactory? Because we are missing something crucial. We really need to look at risk rather than probability alone. Risk is the product of the probability times the impact (almost literally, in this case). We can call this damage parameter D. (The units could for example be the cost of emergency brain surgery).

The amount of damage we can expect in an accident depends on helmet use. With a helmet it is D0, without a helmet it is D1.  Set D0 to 1 for simplicity. We know that D1>>1. For very serious head injuries, which really are the crucial ones, D1 might be 10 or more.

We can then calculate a damage matrix. The calculation is identical for small f.

Screen shot 2014-05-08 at 11.50.40

The values a-d are the damage we can expect within the given time period for that scenario.  To get some grasp if the values, we can set P0=1%, F=0.9, and D1=2 (a very low value).

Screen shot 2014-05-08 at 11.52.29

Clearly, wearing a helmet causes less damage in all scenarios. However, here is the most interesting question: are there any conditions in which a<d, that is, driving voluntarily without a helmet is safer that driving with a mandatory helmet?  We need D1*P0<P0^F, or F < 1+ log(D1)/log(P0). For the sample values above (P0=1%, D1=2) we require that F<85%. If we assume a more realistic D=10, we require F<50%.

Thus, it is possible to envision scenarios in which driving without a helmet is safer. But are these credible scenarios? We would have to assume that mandatory helmets would decrease cycling by tens of percent (even 50%). Possible, but unlikely.

Even more problematic for the author’s case, we would have to assume that the peer pressure of voluntary wearing of helmets would have an effect that is similar to mandatory helmets. Perhaps, but it cannot be as large as the effect of mandatoriness.

There are in fact other arguments against mandatory helmet use. For example, there is a very real phenomenon called the rebound effect. In this case, if safety is improved by a passive solution such as a helmet, then people tend to engage in riskier behaviors because they feel safer doing so. The end result is that safety is not enhanced; it may even be decreased if the perceived improvement is much larger than the actual improvement.

However, this is not really considered in the blog. The core question is: by choosing to cycle without a helmet, is the author significantly increasing the future safety of others, and also by extension himself? Crunching the numbers: no.

Basically, the author is suggesting a massive and highly likely personal sacrifice, for a fairly small and fairly hypothetical improvement. Such a tradeoff is heroic, but it really does not make much sense.

bicycle-crash

Translate »