Thomas John Vitolo
tjvitolo at bu.edu
Tue Mar 8 16:57:55 EST 2005
And here we go...
> > > And if the BPD gets involved, you would have to obey the laws
> such as
> > > traffic lights and
> > > stop signs and so on.
> > Doubtful. In fact, I'd bet what would happen is that the BPD would
> make it more
> > like, ahem, a parade. They'd likely simply block intersections and
> give us a
> > "free pass" without regard to the light. It's far easier and safer
> to keep the
> > group together.
> If you're going to use Engineering arguments, don't use term's like
> "doubtful" and "bet".
Researchers and engineers use "doubtful" and "bet" all the time -- when things
aren't certain. In this particular case, I was talking about what _people_
(namely, the BPD) would do, not an engineering problem anyway. Hence, more
This particular part wasn't an engineering argument, but engineering arguments
ought to contain hedges if the outcome isn't provably 100% certain.
> > > But I think that if more people did turn out, you could have a
> > > lane full of people,
> > > and if they stopped at a light, when that light changed you'd
> > > have a block long of
> > > a road lane full of people, and that would be a good thing.
> > It doesn't work. More specifically, it only could work if
> > (a) the distance between lights was nearly identical, and
> > (b) the light cycles were the same period and the same time for
> that particular
> > street, and
> > (c) the cycles timed out to be exactly the pace of the CM.
> She's talking about filling up the periodic distance between lights
> people. Purely a translational thing.
No -- it's not purely a translational thing. The problem is that once the light
goes red, the other direction gets a go -- and they're cars. They then fill up
the space between light cycles of the CM road.
> If your argument is true
> none of these things happen) then perhaps there is an "optimal sized"
> packet that can travel between blocks that is safe. You probably
> Stoch, right? Maybe you could work on this.
I do know sochastic processes, and I have worked on this. Since Boston doesn't
have city blocks like Manhattan, the optimal sized mass packet -- that is, the
largest size that would have a nearly 0 probability of being segmented, is about
4 or 5 cyclists. Anything more than that and, at the speed CM travels, they'd
likely get caught by a yellow light. This is an estimate, but clearly the size
would be very small. Even smaller if forced to ride single file and in a single
lane. For cars, the number falls to about 3 with average drivers. You can get
it up to 4 or 5 with really skilled (and willing to stretch yellow light)
drivers. A caravan of length longer than this will almost certainly become
segmented in all but very unusual circumstances, such as riding down an avenue
with equally spaced intersections and perfectly timed lights, something Boston
offers very few of.
> Oh, and as most people know
> from commuting, the lights are set up to either help the traffic flow
> impede it. People who actually have their PhD's designed the
Yes, and many people know
(a) Boston isn't like Manhattan, so they can't time the lights properly. The
traffic network simply cannot allow a priority lighting queue for more than a
single road. This is decidedly different than city blocks, where all the
avenues can time the lights, since (i) the distance between intersections is
equal, and (ii) the avenues are parallel. Neither of these exist in Boston,
which means that it's virtually impossible to have two roads anywhere near each
other timed properly.
(b) The traffic infrastructure in Boston is uncommonly old and, well, bad.
(c) The number of cars who run yellow/red lights, block the box, make illegal
left turns, etc. would render the optimization wasted anyway.
(d) People with Ph Ds don't design the systems for the most part. In fact,
that's generally reserved for people with masters degrees, since a Ph D is a
research degree and a Masters an "applied" degree, so to speak.
(e) My masters is in industrial mathmatics, and my thesis involved discretizing
pdes -- a common problem when addressing urban and highway traffic flow.
I've done a fair amount of reading and theoretical work on traffic flow, and
almost pursued a Ph D in Transportation Sciences -- working on exactly these
kinds of problems.
> > This is derived from queueing theory, of which I'm a graduate
> student. Frankly,
> > there's no way to keep the mass together if it stops at lights,
> because even if
> > (a), (b), and (c) were met, people on the cross street turning on
> to the street
> > the CM is on would bisect the stream of cycles. Once you've got a
> > stream, that stream could easily get bisected again at the next
> cycle, etc. For
> > a thought exercise to understand why this is so, imagine a caravan
> of 10 cars
> > driving down Mass ave, all doing exactly the speed limit, and never
> > up/slowing down at lights to stay together. The first car would
> likely get
> > ahead of the last car by over a mile over a 10 mile span.
> Also, that's not true. The probability of hitting a red is the same
> for all of them.
Absolutely false. The probability distribution of a light cycle isn't poisson.
In fact, since the lights are on cycles and the riders are riding at constant
speeds, the probabilities are all either 1 or 0, depending on the rider, and the
correlation between adjacent riders is nearly 1, since if the guy before you
gets a green, you'll almost certainly get a green -- whereas, if the guy a half
block up has green, you'll more likely get red.
> The lead vehical could be at a red while the others
> depending on the velocity, and such. I won't bow to this argument
> until i
> see it worked out, and the burden of proof, if you are going to try
> to be
> academic about it, is on you. So you have a nay-sayer.
Again -- try my thought experiment, or do it for real. Get 5+ vehicles together
(cars or bikes), and agree to travel at the same speed and stay in the same
lane. Agree not to aggressively drive through yellows or be overly cautious
about them. Then, start on Mass Ave way in Cambridge, and drive all the way
through Cambridge and Boston. I guarantee you that the group simply cannot stay
together without "breaking the rules". This can be modeled as a random walk
problem as well.
If I have an hour or two, I'll write up a careful proof that as the number of
intersections increase the probability of a caravan of vehicles remaining
together approaches 0. But, don't hold your breath for it; I'm really busy this
week. Send me an email next week if you're really interested. I'll likely have
much more time then.
> Also please
> included a distribution of wait times based on the route. Either
> way, no
> matter how you model the system, your model is imcomplete and only
> model. Your simple model is rejected soundly based on it's
> simplicity, or as professionals like to say, "It's just a theory".
That's a load of crap. Gravity is a theory -- and a simple, incomplete one at
that. The model is pretty easy since the only objective is to show that a group
simply cannot stay together -- they're certain to get bisected by cars. This
isn't a big leap of faith, either. When the light turns red, some cars on the
cross street will make right/left turns onto the route that CM is on, thereby
placing themselves in between. Now, it's possible that the entire column of
cars will dissipate, bringing the CM together. But, it's also possible that one
of the cars will get stuck at the next red light, and even more cars will enter
from the sidestreet. Now, the distance that the CM is bisected (measured in
carlengths) is even larger.
This phenomenon can also explain why the bus is "always late." You see, it
simply isn't possible for it to be early. It can eithe be on time (it can't
leave a stop before its scheduled time), or late. If it's late, it can either
catch up or end up even later. If it ends up even later, it can either catch up
somewhat or end up even later than that. Etc.
The biks in the back of CM can't, by definition, make the distance between them
and the front of CM negative -- it can be 0 or positive (have cars between it).
So, eventually it'll get cars between it, and cars between that.
> > > And
> > > still, the car traffic
> > > could get by, and people could see the beauty of what CM can be.
> > > I think it is just such
> > > a shame that it isn't that now, and as CM keeps to the marginal
> > > fringes, that margin is
> > > going to get narrower andnarrower, which is such a shame.
The traffic could get by, but at severe costs to the density and safety of the
CM -- a tradeoff that doesn't seem worth it to me. As I wrote earlier:
> > Ultimately, since I believe that stopping at lights fractures CM in
> a way that
> > is both far less safe and less effective for CM, I believe that
> riding in 1 lane
> > is also a show-stopper, since it leads to (a) cars trying to make
> right or left
> > turns through the mass, (b) cars speeding to pass the mass (which
> can be very
> > dangerous to the cyclists and any other motorists/peds in the
> area), and (c)
> > added confusion when the CM doesn't stop at a red light, since it's
> "half" or
> > "one third" as wide, and therefore "half" or "one third" as massive
> from the
> > perspective of another set of vehicles preparing to go through the
> newly green
> > light.
> Hey, if people followed traffic laws (all people) this wouldn't be
Yes it would. There'd still be cars merging left or right into the CM to turn
at the next intersection. That's a completely legal automobile manuver which
would negatively impact the CM's safety and density.
> > In my opinion, CM simply can't function safely or effectively if
> > fragmented by traffic lights, or if it allows cars to pass by
> > in a single lane. It's best method of keeping cars from growing
> > frustrated, if that is important, is to
> > not ride along the same road for very long; constant "wiggling"
> between streets
> > and avenues will dramatically reduce the average and worst case
> waiting times
> > for motor vehicles, at the "cost" of distributing the delay amongst
> many more
> > drivers. Additionally, not riding right at 5:00pm (which the
> Boston CM doesn't)
> > helps tremendously. The 6:30ish ride is at the back half of rush
> hour, on the
> > day of the week when many more people are likely to leave work
> You need to rewrite this entire conclusion. You make wild claims
> substantiation, highly unacademic of you.
I haven't bothered to write proofs or cite articles in this unacademic forum.
However, the claims aren't wild, nor are the conclusions without substantiation.
I made clear arguments about
* how CM would end up splintered due to traffic lights
* how legal and illegal motorcar manuvers would put CM at risk
* how altering the route of CM to "wiggle" would keep it off of (nearly) all
commuters path for long distances of time, thereby clearly reducing the worst
case delay for motor vehicles.
The arguments are simple and straightforward. You can feel free to argue that
everything I write regarding an academic subject should be held to the highest
journal standards, but most folks will ignore that argument. The goal was to
express simple ideas from a complex system, not to further the state of the art
theoretical understanding of traffic engineering.
> Why are you tossing around terms without defining them?
I felt reasonably intelligent people on this list would understand the definition.
Yeah -- cost. You know, the penalty you incur. The cost, or "cost" as I put it
to indicate my looseness with the definition, is the added length of time a
commuter faces due to CM. Exactly what we've been talking about.
> Worst case?
The worst case. As in, you take all the cases (of added automobile commuted
length, since that's what we're talking about), and you find the one which added
the most time (since we're claiming that a longer commute is worse). That's the
> Where's your trendline young man!
> ps. Half of this is tongue and cheek,
(like my above comment)
> the other half serious. Just
> because you study when to shift bits, doesn't make you an expert on
> traffic. It's a common misconception to beieve that, if we are an
> in one thing, we must be an expert in other things.
Yeah -- and it's also a common misconception that if one is an expert in one
thing, he can't be in other things. The fact is that I have studied traffic
engineering, from a civil engineering perspective, a stochastic processes
perspective, and a network engineering perspective. Am I an expert? Well -- I
guess that depends on the company I'm currently keeping. I'd make the assertion
that I am an expert on traffic flow on this mailing list. I might be wrong; for
all I know, half of MIT's CTS group is on this mailing list. But, I'd bet
(there's that word again) that isn't the case.
And BTW, I'm not a "bit shifter" -- I'm not sure what exactly you think I study.
Systems Engineering doesn't refer solely to computer systems... it refers to
all systems, such as manufacturing systems, network-queueing systems, and...
wait for it... traffic systems. The research I'm working on right now (mdMST
problems) are highly relevant in urban planning, although that's not the
particular application I'm using.
Thomas John Vitolo
Ph D Systems Engineering Candidate,
Support a few technologists in Washington. Go to:
More information about the Bostoncriticalmass