Saturday, October 25, 2014

Hit Me Not In the Darkening Dusk

Do you see the bicycle in this photo? Distance just over 100 feet, at dusk, Philips LED Bike light on econ setting

As it became clear to me that the ultimate questions related to not being hit while commuting by bike at dusk are binary--do drivers see me yes/no? do I avoid being hit by drivers yes/no?--so too the complexity of the inputs to those questions became apparent. After experiencing not one but two near-NO answers to the second question on Friday's commute, I wanted to mull over those inputs more, to support a reasoned and effective approach for the darkening commutes ahead while still striking a sensible, achievable, and acceptable balance.

The questions that began to pile up and which require at least working hypotheses to support that reasoned and effective approach (in no particular order):

  • Do drivers see me?
  • Do drivers react in time?
  • Do I avoid being hit by drivers?
  • How do ambient lighting conditions factor in? Including sunlight brightness and angle, street lights, other lights, background, rain/fog/dust conditions, background and foreground conditions?
  • Does my lighting dazzle or irritate other road users in an impolite or hazardous manner?
  • How do the angles of incidence factor in, particularly for side-entering vehicles?
  • How do you catch the attention of drivers in time to facilitate the not-getting-hit result?
  • How does perceived brightness factor in?
  • How do beam shape and beam cutoff factor in?
  • Doesn't the usage by other road users of very bright and dazzling lights establish a standard of reasonable use? (search for H7, HIR2, H9, H11 bulbs for details)
  • Flashing or steady? What flash pattern? Single or multiple lights? What configuration?
  • How does light height and aim factor in?
  • How much variability is there in the driver-centric factors? 
  • Oh and is the solution, practical, affordable, secure, and reliable?

Several of these are non-linear functions of multiple variables, which makes the assessment more difficult. In addition, most or all of them are at their roots somewhat controversial or undecided questions of human psychology, perception, physics, engineering, optics, and so on. In short, there's even more uncertainty behind the uncertainty. Where to begin?

Let's dive in with what may be the more complex factors. The perceived brightness of lights, catching the attention of drivers, probably even the perceived level of politeness of any given act, are not simple, linear functions, but seem to be most accurately modeled by the Stevens' Power Law. That says that the results are not a straight one-for-one with the inputs, but instead have a relationship that involves a constant and an exponent: a one unit increase in the output of a light does not result in a one unit increase in the perceived brightness, but rather less. Similarly, catching the attention of a driver who doesn't see you may require a lot more input energy on your part than you may think, since a one unit increase on your part does not accomplish a one unit increase of attention on their part, but rather less. Lastly, while it may seem on your part that you are being twice as polite on the manners scale, they may perceive you as only being one unit more polite.

On top of that, it's not only a specific reaction in a driver's brain that we are shooting for here, but more importantly an action on their part, fast enough, as a result of that specific brain reaction: don't hit the cyclist! Human decision and action also is the result of complex and non-linear functions which are beyond the current knowledge level of the OSG Institute of Behavior Studies, perhaps at a detailed level beyond the present understanding of science. 

However, observationally, and as I mentioned in the beginning, at possibly a level of understanding which supports an effective decision even if all the details are not yet understood, I find the activation energy model of human decision and behavior useful. This is actually behind the tagline of this blog from the beginning: get up, go ride. There's something like inertia which hinders human decision and action, and this inertia requires more input energy to overcome, for less result, than our intuition tells us. How hard is it really to just get up off the couch? How difficult really is it to just get on the bicycle and ride? Answer: more than just the simple sum of the efforts of standing up plus riding the bike. In practice, the effort required feels more like the graph of chemical activation energy:

Activation energy path graph, CC-licensed image source: Wikimedia

What I'm trying to represent with this graph is that it requires more input energy to cause action that we might think: something has to push you over the hump to cause action and get you to the bike riding level of energy, whether it's a significant input in energy, or lowering the hump itself with a catalyst. If the input energy is the perceived brightness of a bicycle light, which is itself a Stevens' Power Law function, even the basic question about how bright a light is needed to trigger the desired action in a driver can be seen as complex. 

Perception[a Stevens' Power Law function] --triggers--> action[reaction path function of human motivation]

Add in that the result is also subject to driver reaction time, driver decision process, and driver physical abilities. I've written about a simplified version of these factors before, that I called "dirty space": at 45mph, the common speed of vehicles on several of the streets on my commute, the space in front of a vehicle which is required by the combination of all these factors to be avoided in order to give the driver enough time and space to react, act, and avoid, is somewhere between 100 and 200 feet, and could be more depending on other factors (fog, rain, dirty windshield, impairment, distraction, etc).

For the question of avoiding being hit in the darkening dusk, this dirty space means that the requirements on any proposed solution have the additional (related, velocity-dependent) challenges of distance and time: it needs to be effective at at least one hundred and better two hundred feet, and it needs to work in the time frame of one to three seconds. Shorter than that in distance, or slower than that in time, and you might as well not have any lights at all.

Also related, and going to come out one way or another depending on the solution, are the question of politeness, dazzling, beam cutoff, and safety. 

From a similar post three years ago, the very bright Tri-Newt approach

Trying to work out all these factors is daunting. Yet, a light must be chosen and used. Considered together in combination, we may have so many variables and unknowns that the most effective practical solution is the engineer's: run your calculations, recognize that there are many unknowns, dependent variables, and some wide error bars, and err on the side of safety. 

Three years ago, after a seemingly neverending series of near SMIDSY almost-hit-mes at dusk or evening, I starting using my Niterider Trinewt all the time. That's the "normal" setting in the photo. On the plus side, everyone saw me. No more near hit-mes, ever. On the downside, drivers all the way across the road flashed their brights at me several times, and one or two yelled at me.

I'll go ahead and say, based on those observations, that the Trinewt probably exceeded the ideal brightness and/or lacked a sufficiently cutoff beam shape to avoid pissing off drivers. If I could have anything I wanted, it would be an automatically activated Trinewt that flashed on high setting whenever a side-entering driver was detected in an about-to-hit-me situation.*

So, let's say I choose a main light that is more polite than that, both in brightness, and in having a cutoff beam shape that is supposed to be more driver-friendly. Would that work? That's what I've got in the first photo at the top of this post: my Philips Saferide LED bike light, set on economy (or normal) mode. Maybe I need to try running it on the bright setting--the beam cutoff should still help, right? Or, maybe I need to augment it with something like this on my helmet:

Knog "hipster cyst" flasher attached to front of helmet experimentally

When you get down to it, this is a question of balance: is it even possible to balance visibility with politeness? The Knog attached to the helmet in an always-on always-flashing state has potential to push it back over to the impolite dazzling light side. This configuration has the benefit of flashing where I look, which does alert side-entering drivers very effectively. Way back in the beginning of my bicycle commuting odyssey, I went a little bit nuts with 3M dual-lock tape and both Inova microlights as well as Photon Micro-light 3s, sticking them everywhere particularly on my helmet in a flash-where-I-look configuration. They seemed to work for the intended purpose, did not seem impolitely bright, and were affordable and reliable. If anything, they fell down a bit on the fiddle factor scale: micro-lights dual-locked to your helmet does not even have the sound of a finely engineered solution to it.**

So, in conclusion, what about this Fall, what am I going with? Running the Philips with it's shaped beam and sharp cutoff on bright (particularly during the darkening dusk) is going to run down the batteries faster, but they are rechargeable, and I carry backup lights. I'm going to give that a try. Also, I'm charging up the Trinewt. Just in case. Because I'm not sure, given all the variables and unknowns, that it actually is possible to strike or find the ideal balance between safety and civility: don't get hit, don't make anyone mad with a light that's too bright. If I have to edge into the second in order to achieve the first, I think that's OK. My intentions are good. In the binary reality of seen/not seen, not-hit/hit, I'll do what's necessary to support the not-hit result to the best of my understanding of a complex situation and like an engineer, err on the side of safety. 

*On a side note, I acknowledge the sick sense of humor that gave the Trinewt a fast strobe setting which lacks any appropriate usage but is freaking awesome. Its switch, though, in its out-of-the-box state, does not appear to be easily modifiable to something like a handlebar-mounted panic switch. Once more into the hacking box, dear friends.

** I still carry a Fenix P1D powered by a rechargeable RCR123 with dual-lock strip affixed just in case. A solution which involves attaching anything to a bike helmet, though, lacks elegance, and possibly compromises safety.   


  1. I am always vigilant when traveling south on Hayden, turning right on Indian Bend which I do every workday. That walkway/bikeway thing there that is semi-enclosed is so hazardous, I think. It's really hard to tell if there is a bike in there that is going to go through the green light while I'm turning. Poor design feature, I think. I can see bikes in the open but I can't necessarily see them well enclosed.

    1. Yes! Just down the way where it crosses N. 78 Place gives me pause every time I cross--you can't really see if any cars are coming, and there's no real indication of signage or paint to tell you who should do what. At dusk, I imagine that's one of the side-entering scenarios that a bright light that throws a wider pattern with some side lobes might help. I'm all for slowing down for safety. That's a good example of somehow needing to let them know you're there.


Please feel free to comment here, almost anything goes, except for obvious spam or blatantly illegal or objectionable material. Spammers may be subject to public ridicule, scorn, or outright shaming, and the companies represented in spam shall earn disrepute and ire for each occurrence.