The article contains a link to a YouTube video of the Loveparade 2010. It's pretty long and only at the end the oscillations are very noticeable (18:33 [0]).
21 people died and at least 652 were in part seriously harmed.
Crowds of a high enough density push people in orbital waves, with both clockwise and counterclockwise oscillations pulsing at 18 second intervals through the crowd. Monitoring these motions in realtime is possible and may help prevent crowd crush events.
Also: At high densities crowds act like sticky springs, with the ground adding the sticky.
18 seconds before the next wave of people crushes you. 18 seconds to get out. 18 seconds could save your life.
alwa 34 days ago [-]
I read it kind of quickly, but didn’t their analysis suggest that the 18s period in question may have been specific to the Plaza Consistorial—that is, that the specific oscillation period may depend on the size (and possibly shape) of the confinement area?
> To address the role of confinement, we perform an additional series of measurements after the festival opening, when a security team splits the crowd into two halves (Fig. 3e and Supplementary Video 4). We find that the two decoupled crowds still oscillate but at a higher frequency. We measure the shortest dimension of the crowds before and after the festival opening, and the extent of the region of space where the 2010 Love Parade crowd featured chiral oscillations (Methods). Figure 3h shows that the velocity spectra of these markedly different crowds peak at the same value when rescaling the frequency w by the inverse of the confinement length L. This result strongly suggests that the period of the emergent oscillations depends on confinement, and is not an intrinsic 'material' property.
cmehdy 35 days ago [-]
Two things that would deserve clarification (although outside of the scope of the research here) are:
(1) When is it time to leave? At what threshold? Is it around an arm's length of free space between people?
And (2) Where to go? Straight line towards an exit? Perpendicular to the crowd? Exactly against the flow? Along the flow but towards edges? Probably depends on a lot of parameters..
wongarsu 34 days ago [-]
A naive application of fluid dynamics would suggest that the answer to 2 is to go perpendicular to the flow until you reach a wall, then go against the flow towards an exit.
In the middle of a crowd/fluid your motion is determined by the humans/molecules around you. The closer you get to a wall, the fewer particles have an influence on you. Just make sure it's not a wall people are moving towards.
Not sure if anyone has studied how well this holds up to humans. Human crowds have very fluid-like behavior, but of course they don't behave perfectly like a liquid in a pipe
alwa 34 days ago [-]
Is there anything to the notion that you might prefer to be batted around in the “soft body” portion of the dense-crowd “fluid” rather than pinned between several tons of soft bodies in motion and a hard limit? You seem rather more crushable than an individual molecule might be…
(edit: “crowd crush safety expert” Paul Wertheimer, who came to some notoriety through studying mosh pits from within them, apparently suggests—at least in situations where the crowd is moving toward a clear focal point—you’re right: he recommends the edges
That said I note that the authors of TFA draw a distinction between dense-crowd situations undergoing unidirectional flow and the steady, confined dense crowd whose behavior they describe.)
xeonmc 34 days ago [-]
> Human crowds have very fluid-like behavior, but of course they don't behave perfectly like a liquid in a pipe
Hypothetically, how would the dynamic change if it were a crowd of cats instead?
thfuran 34 days ago [-]
Non-cat fluids don't usually spontaneously draw knives and start cutting themselves apart.
vxxzy 35 days ago [-]
I don’t recall where, but for some reason I remember learning that if people start to bump into you too many times, that’s a warning shot to get out. It was something like “bumps per second”.
m463 34 days ago [-]
I wonder if we couldn't have the equivalent of the water barrels near the start of guard rails, but for crowds?
Maybe giant inflated or crushable areas around an exit that could take up slack?
maybe tie the crush event with something else, like opening extra doors, turning on lighted warning signs or illuminating exits?
lifeformed 34 days ago [-]
Doors with crash bars are kind of that. If there's too many people, they'll push up against them and force them open.
metalman 34 days ago [-]
.,.......except at a Greatfull Dead concert where this phenominon is called the "seeing the gears"
or "have you seen the gears"
nobody gets crushed, everybody gets....where they need to be
plus up in the stands, are human waves of people, who are also bouncing a multitude of beach balls around, some of which are giant, some of which are not filled with air
everybody steping to the groove.....except the spinners, who are another phenominon, the tapers huddled, but looking for any likely person at all to spell them
bobby just played the grammys......
patcon 34 days ago [-]
Of note: If that 18 second rule became known to a significant number of people, the dynamic will change
trimethylpurine 35 days ago [-]
I like your summary. Thanks!
diegoperini 35 days ago [-]
Beautiful. Thank you.
samstave 34 days ago [-]
Id bet that the 18 seconds is related to the amount of time between when people naturally shift on their feet/comfort pocket...
Meaning that, if you notice the frequency by which people shift their weight and move -- its likely in that window, assuming other externalities (bathroom, heat, thirst, anger, etc -- the physiological cycle of body comfort may be in that area of 18 secs)
paulorlando 34 days ago [-]
I used to have a view from about 3 stories up of one of the densest crosswalks in Hong Kong. As the crowd of people crossed the street in two directions, they would organize into 7 rivers of flow (3 one way, 4 in the other). The rivers were surprisingly consistent over the months I watched this.
kanavs 35 days ago [-]
Interesting work! It is interesting how a lot of the models in physics and outside of it comes down to variations of harmonic oscillators (in this case macroscopic chiral oscillators). I am curious is the chirality gets imposed by the location and the architecture. Another interesting bit to explore would be how the agency (in my head agency is loosely defined as the number of possible actions you can take) of each individual gets contained as the crowd gets dense. If there is a way to define agency mathematically, it would be interesting to look at the correlation between agency and crowd density (my guess is inverse correlation) and then come up with strategies to increase the agency.
alwa 34 days ago [-]
What cool work! I wish I had this transformative sensation more often: of the quick, clear shift in my thinking after someone’s careful study of a familiar problem results in convincing models simple enough for a non-expert to wrap their mind around.
If this interests you, some prior context on “crowd crush” might too. John Seabrook, writing in the New Yorker in 2011 (in response to the “Black Friday” crushing deaths at US Wal-Marts), engagingly profiled examples from the past few decades of this category of disaster (in the US and Europe). Along the way he traces a few shifts in thinking about such crowd behavior, from “mob psychology” to Paul Wertheimer‘s colorful observational studies to Dirk Helbing’s fluid-dynamical modeling.
I recommend William Bialek's lectures who deep dives into how maths can be applied to messy macro-world of biology, neurons, cells, birds, so on and so forth.
Precision and emergence in the physics of life with William Bialek
All systems with finite speed of information propagation will produce oscillatory phenomena emergently.
Nature is short-sighted <=> Nature is governed by differential equations/local behaviour
TeMPOraL 34 days ago [-]
Now I wonder if there's some insight to be gained from looking at the crowd behavior in the frequency domain, i.e. through a Fourier transform?
deadbabe 34 days ago [-]
Most dangerous crowd I’ve been in was in 2021 at Astroworld, several people died from being crushed, it felt like there was no escape.
silver911r 35 days ago [-]
I noticed a similar phenomenon while driving on the highway. Vehicles on the other side of the road seem to group up and travel in clusters. Like traffic naturally oscillates into these packets?
jrootabega 34 days ago [-]
The traffic waves mentioned in the other replies are not the same as clumping. People clump up while driving, even on the emptiest of highways, because they are bad drivers. This was happening long before automatic braking, but that has made it more common.
taurknaut 34 days ago [-]
How is clumping indicative of bad driving?
jrootabega 34 days ago [-]
Unnecessary clumps are bad. They make merging and passing harder, desensitize you to nearby cars' presence, cause night vision problems, etc.
A good driver will first notice they are in an unnecessary clump, and a bad driver probably won't.
A good driver will try to minimize the the time they spend relatively motionless to other cars (other than behind them at a safe follow distance). A bad driver either doesn't realize as in the previous point, doesn't care, or actually prefers it as some kind of crutch. Sometimes the leader of the clump ends up that way because they will speed up whenever those behind them try to pass. Sometimes the rear of the pack is a driving zombie who will close any distance between them and the car in front, but will never choose to pass. And so on...
therein 34 days ago [-]
In the Bay area, what I noticed is, unnecessary clumps almost always get nucleated by a Tesla going lower than the speed limit on the left lane and then a tall vehicle behind them.
The tall vehicle behind the Tesla will hide the fact that the slow Tesla on the left lane is slow for no reason, people behind the tall vehicle will think there is traffic. And we will all go 55 on the left lane, thinking there is traffic. Until some incline allows people to see that there is no real traffic and then the Tesla gets overtaken and clump is broken.
greentxt 34 days ago [-]
Stay in the far right lane unless passing. Don't tailgate. Those 2 rules would solve a lot of problems if they were enforced. Instead, there is heavy enforcement of the speed limit, which in and of itself isn't a safety concern.
greentxt 34 days ago [-]
Left lanes are for passing. If you are in the left lane and not passing, guess what.
jrootabega 34 days ago [-]
I agree that clumps usually end up blocking passing lanes, but they're also bad when in any lane(s).
greentxt 34 days ago [-]
A clump == blocking the passing lane. They are definitionally the same.
taurknaut 34 days ago [-]
I think about a large cluster of cars, say 6-12 cars, going roughly the same speed. Nothing about that entails blocking multiple lanes, certainly not densely enough to preclude someone from passing.
pglevy 34 days ago [-]
Guess what... Sometimes I need to take a left exit. And there will frequently be a bunch of clumping chumps doing 20 over in the left lane. So I get over when I can.
tristor 34 days ago [-]
Because it indicates that they are not driving to the conditions of the road and being attentive.
jampekka 35 days ago [-]
This is a well known phenomenon in traffic engineering and can be modeled with e.g. fluid dynamics.
I had heard of traffic waves in congestion, but this phenomenon happens on open roads. I often notice vehicles forming packets when there isn't traffic. Like natural clustering? If I drive the speed limit, cars will start to bunch up behind me, even when there is an open lane. Then, a driver will speed up in the passing lane, but when they reach me, they often slow down slightly or a lot and ride next to me without passing.
lubujackson 34 days ago [-]
There is very specific clumping behavior on the I5 in Northern California I've noticed. It is a long, 2 lane highway often used by truckers. It tends to be fairly congested but not ever backed up.
What I see every time I drive it is a new clumping behavior that emerges: some big rigs tend to drive quite slow but do that thing where they want to pass someone going 54 and they are going 55 so it takes ages and the traffic in the left lane gets longer and longer, leaving a gap in the right lane.
At thing point, some drivers say "fuck it" and zip down the right lane and aggressively merge back in to shorten their wait getting around the blockage. As you are likely driving this route for hours (SF to LA) you see it happen more and more often and it becomes more and more prevalent. I find it exhausting because there is no way to veg the whole drive unless you are the slowest car out there because at best someone is going to jab in front of you at the last second and you have to mash you brakes to avoid them.
The interesting thing is that this is always how the traffic flows on this road, and nowhere else. It must be a precise combination of traffic levels, lanes available, traffic mix, differring speeds and even average distance travelling, so people notice and feel compelled to respond to the pattern.
jrootabega 34 days ago [-]
In my experience, anywhere there is heavy semi truck traffic will have significant problems once there are enough cars around ("enough" being far less than it should be). 2 lanes, 3 lanes, 4 lanes, hills, it's all bad. But yeah, only 2 lanes per direction makes it comically awful. Happens on sections of I-81 too.
I think the specific pattern you're talking about happens because when a truck is passing slowly but infrequently, it makes some sense to be polite and line up behind it. When trucks are passing so frequently, and so incredibly slowly, being polite is useless. There needs to be a zipper merge happening behind the truck being passed. But the left lane people feel they've paid their dues by waiting in line, and half the right lane people don't even care. The right lane people also stay so close to the truck in front that they can't safely merge left at highway speeds anyway. People also tend to not fill back in to the right lane between truck passes. I'm not sure that would even help, but it does make people who do seem like cheaters to the people who refuse to.
21 people died and at least 652 were in part seriously harmed.
[0] https://youtu.be/QpzISdBE3dA?t=1113
18 seconds before the next wave of people crushes you. 18 seconds to get out. 18 seconds could save your life.
> To address the role of confinement, we perform an additional series of measurements after the festival opening, when a security team splits the crowd into two halves (Fig. 3e and Supplementary Video 4). We find that the two decoupled crowds still oscillate but at a higher frequency. We measure the shortest dimension of the crowds before and after the festival opening, and the extent of the region of space where the 2010 Love Parade crowd featured chiral oscillations (Methods). Figure 3h shows that the velocity spectra of these markedly different crowds peak at the same value when rescaling the frequency w by the inverse of the confinement length L. This result strongly suggests that the period of the emergent oscillations depends on confinement, and is not an intrinsic 'material' property.
(1) When is it time to leave? At what threshold? Is it around an arm's length of free space between people?
And (2) Where to go? Straight line towards an exit? Perpendicular to the crowd? Exactly against the flow? Along the flow but towards edges? Probably depends on a lot of parameters..
In the middle of a crowd/fluid your motion is determined by the humans/molecules around you. The closer you get to a wall, the fewer particles have an influence on you. Just make sure it's not a wall people are moving towards.
Not sure if anyone has studied how well this holds up to humans. Human crowds have very fluid-like behavior, but of course they don't behave perfectly like a liquid in a pipe
(edit: “crowd crush safety expert” Paul Wertheimer, who came to some notoriety through studying mosh pits from within them, apparently suggests—at least in situations where the crowd is moving toward a clear focal point—you’re right: he recommends the edges
https://www.cnn.com/2021/11/13/us/how-to-stay-safe-at-concer...
That said I note that the authors of TFA draw a distinction between dense-crowd situations undergoing unidirectional flow and the steady, confined dense crowd whose behavior they describe.)
Hypothetically, how would the dynamic change if it were a crowd of cats instead?
Maybe giant inflated or crushable areas around an exit that could take up slack?
maybe tie the crush event with something else, like opening extra doors, turning on lighted warning signs or illuminating exits?
Meaning that, if you notice the frequency by which people shift their weight and move -- its likely in that window, assuming other externalities (bathroom, heat, thirst, anger, etc -- the physiological cycle of body comfort may be in that area of 18 secs)
If this interests you, some prior context on “crowd crush” might too. John Seabrook, writing in the New Yorker in 2011 (in response to the “Black Friday” crushing deaths at US Wal-Marts), engagingly profiled examples from the past few decades of this category of disaster (in the US and Europe). Along the way he traces a few shifts in thinking about such crowd behavior, from “mob psychology” to Paul Wertheimer‘s colorful observational studies to Dirk Helbing’s fluid-dynamical modeling.
https://www.newyorker.com/magazine/2011/02/07/crush-point
https://archive.is/b1Flz
Precision and emergence in the physics of life with William Bialek
https://www.youtube.com/watch?v=c1hd5El2gqc
Nature is short-sighted <=> Nature is governed by differential equations/local behaviour
A good driver will first notice they are in an unnecessary clump, and a bad driver probably won't.
A good driver will try to minimize the the time they spend relatively motionless to other cars (other than behind them at a safe follow distance). A bad driver either doesn't realize as in the previous point, doesn't care, or actually prefers it as some kind of crutch. Sometimes the leader of the clump ends up that way because they will speed up whenever those behind them try to pass. Sometimes the rear of the pack is a driving zombie who will close any distance between them and the car in front, but will never choose to pass. And so on...
The tall vehicle behind the Tesla will hide the fact that the slow Tesla on the left lane is slow for no reason, people behind the tall vehicle will think there is traffic. And we will all go 55 on the left lane, thinking there is traffic. Until some incline allows people to see that there is no real traffic and then the Tesla gets overtaken and clump is broken.
https://en.m.wikipedia.org/wiki/Traffic_wave
What I see every time I drive it is a new clumping behavior that emerges: some big rigs tend to drive quite slow but do that thing where they want to pass someone going 54 and they are going 55 so it takes ages and the traffic in the left lane gets longer and longer, leaving a gap in the right lane.
At thing point, some drivers say "fuck it" and zip down the right lane and aggressively merge back in to shorten their wait getting around the blockage. As you are likely driving this route for hours (SF to LA) you see it happen more and more often and it becomes more and more prevalent. I find it exhausting because there is no way to veg the whole drive unless you are the slowest car out there because at best someone is going to jab in front of you at the last second and you have to mash you brakes to avoid them.
The interesting thing is that this is always how the traffic flows on this road, and nowhere else. It must be a precise combination of traffic levels, lanes available, traffic mix, differring speeds and even average distance travelling, so people notice and feel compelled to respond to the pattern.
I think the specific pattern you're talking about happens because when a truck is passing slowly but infrequently, it makes some sense to be polite and line up behind it. When trucks are passing so frequently, and so incredibly slowly, being polite is useless. There needs to be a zipper merge happening behind the truck being passed. But the left lane people feel they've paid their dues by waiting in line, and half the right lane people don't even care. The right lane people also stay so close to the truck in front that they can't safely merge left at highway speeds anyway. People also tend to not fill back in to the right lane between truck passes. I'm not sure that would even help, but it does make people who do seem like cheaters to the people who refuse to.
Just to add, this paper has nice figures generally!