271 - pyFDS-Evac with Mohcine Chraibi

Why Couple Fire And Evacuation

Wojciech Wegrzynski 0:00

Hello, everybody. Welcome to the Fire Science Show. As fire engineers, we do a lot of fire modeling, but we also do a lot of evacuation modeling. Humans are important to us, and in many projects, they are as important as the smoke field. Uh, what's best? Best is to combine the both of the worlds and run, uh, evacuation simulations and the fire simulations at the same time. Uh, if you're old enough, well, you don't have to be that old, maybe that sounds bad, but, uh, if you, if you're in the field for some years, you may remember FDS ha- used to have this capacity to do simple evacuation simulations through a model called Evac. It, it's been there. It's been removed since then from FDS. We don't have it anymore. when Evac disappeared, there was kind of a gap, you know, to, to do this, uh, ty- type of joint, uh, analysis. Eventually, commercial softwares like, Pathfinder caught up, and now for the world of, professional engineering, we, we have this. But, while I appreciate those tools, and I, I like them, and they, they're great and they're very useful, and they have beautiful post-processing, et cetera, et cetera, they're not research tools. They're... Like, it's very difficult to modify them without breaking them, you know? And they, they, they have-- While they are documented, th- there's always this, uh, hidden assumptions behind them. Therefore, I highly support developing, uh, open tools for researchers, uh, in mind and such a tool we're discussing today is called pyFDS Evac and it's developed at Jülich Forschungszentrum. Today I am interviewing, uh, Mohcine Chraibi who is a senior researcher at Forschungszentrum and he's focusing on pedestrian dynamics, crowd behaviors and one of the developers of, of this magnificent package which, uh, is supposed to streamline the, the flow from FDS to, to human evacuation simulations and, and allow exactly what Evac allowed years ago but now in a completely new framework using JuPedSim as its, uh, human, uh, movement model, interesting pathfinding solutions and, and all the integrations you need to, to actually carry out, uh, coupled simulations, uh, of humans and, and the fire field. Very interesting developments. We are sitting down at Wuppertal. We've just finished up the Visibility in Smoke workshop, uh, at Wuppertal. It was fantastic, and more on that will come later on. In that spirit, I just took Mochin and, and did this interview. I've enjoyed it thoroughly, and I hope so will you. Uh, important thing, links are in the show notes. They're always in the show notes, but I really have to emphasize it right now because, uh, we're talking about an open package. You can just download it right now and play with it as you listen to the podcast episode if you like that. Uh, if you just like to listen, that's also perfectly fine, and I welcome you to do that. Let's spin the intro and jump into the episode.

Sponsor Message From OFR Consultants

Wojciech Wegrzynski 3:20

The Fire Science Show podcast is brought to you in collaboration with OFR Consultants, a multi-award-winning independent consultancy dedicated to addressing fire safety challenges. OFR is the UK's leading fire risk consultancy that this year celebrates its 10th anniversary. As experts in fire engineering, they are fully committed to delivering preeminent expertise to protect people, property, and the environment. With over 30 chartered engineers and a team of fire researchers at their core, they continually explore the challenges that fire creates for their clients and society so that the best research, experience, and diligence can be applied for effective tailored solutions. In 2026, OFR will grow its team once again and is keen to hear from industry professionals who want to collaborate on fire safety features this year. Get in touch at ofrconsultants.com. And now back to the episode

Meet Mohcine And The Jülich Team

Wojciech Wegrzynski 4:21

Hello, everybody. I am joined today by Mohcine Chraibi from Jülich Forschungszentrum. Hey, Mohcine.

Mohcine Chraibi 4:26

Hi, Wojciech.

Wojciech Wegrzynski 4:28

Good, good, good to have you in the podcast. I stole you straight from the room where you just delivered the presentation. So as, as things are fresh in my mind and, uh, and you have, uh, warmed up, let's, let's do this for a bigger audience.

Mohcine Chraibi 4:42

Which workshop? The visibility workshop in Wuppertal?

Wojciech Wegrzynski 4:44

Yes.

Mohcine Chraibi 4:45

Yes, and luckily we had lunch before we can-

Wojciech Wegrzynski 4:47

that's, uh... Well, that's, that means we can go for hours. Mohsin, you've shown the developments of, uh, evacuation tool, JJuPedSim, which is, uh, something that people can access already, uh, through the website, which is obviously in the show notes of the episode. And I would like to hear the story of, how it came to life and, why does your team develop this, and then what do you intend to do with it? And, and maybe we, we should also introduce in broader the, the, the human evacuation part of Jülich, Forschungszentrum because that's a very big group. I, I see a picture behind you of, uh, of people moving through boxes, so. Yes. Uh,

Mohcine Chraibi 5:28

So, so Jülich, in Jülich, we... In our institutes, we have, At Civil Safety, uh, Research Institute, we have four divisions.

Wojciech Wegrzynski 5:35

Mm-hmm.

Mohcine Chraibi 5:36

three of them are pedestrian dynamics related, and one of them is fire simulations related. Uh, so you have, Lukas' division, they do fire simulations. And then you have my division, for example, we do simulations and modeling.

Wojciech Wegrzynski 5:50

Okay.

Mohcine Chraibi 5:51

This is the theory. And then there is, Mike's division, they do experiments.

Wojciech Wegrzynski 5:55

Mm-hmm.

Mohcine Chraibi 5:55

Maybe you've seen the posters. So they do a lot of experiments.

Wojciech Wegrzynski 5:59

People with arrows on their heads, uh, yeah. Exactly

Mohcine Chraibi 6:00

those guys. And they collect the data, and then they collect the trajectories and do the analysis of them.

Wojciech Wegrzynski 6:05

Mm-hmm.

Mohcine Chraibi 6:06

And then there is the, the fourth division is, uh, psycho-so-sociology. From Anna. And this, so this is the, division that is supposed to bring the human element to our model, actually. So, ultimately, if all divisions' work, is combined, then you end up with human behavior and fire, as we talked before.

Wojciech Wegrzynski 6:26

Yes.

Mohcine Chraibi 6:26

Which is a, a big thing and, uh, not so easy to, to defy, Wojciech

Wojciech Wegrzynski 6:32

now, uh, in terms of developing the, the Jupet, you said y- you said you, you, you do modeling.

Mohcine Chraibi 6:38

Yes.

Wojciech Wegrzynski 6:38

Uh, so, so what did you use before and what was the gap did you try to fill with this software?

How JuPedSim Outgrew Its Codebase

Mohcine Chraibi 6:43

So JJuPedSim, I think it started, like, around 2008.

Wojciech Wegrzynski 6:47

Mm-hmm.

Mohcine Chraibi 6:47

At the time we had a project where, uh, three PhD students, they were working on their PhD s- thesis actually. Uh, one was doing routing and way finding. The other one was me doing models-

Wojciech Wegrzynski 6:59

Mm

Mohcine Chraibi 6:59

on operational level. And then the third guy was doing analysis of data. And, eventually everyone finished, and then we thought, "Okay, so, um, what can we do? What should we do with this? Let's just put it all together and- Mm call it JJuPedSim." Okay? So it's a big mess of, uh, spaghetti code and, it's working, but it's, At the time it was fine-

Wojciech Wegrzynski 7:20

Mm

Mohcine Chraibi 7:21

for us. But then PhD students keep, coming, implementing their stuff, and going. Okay. You know? So you can imagine at some point it's not maintainable anymore, so. Okay. That was where we decided five years ago that, um, we can't go on forever like this.

Wojciech Wegrzynski 7:37

Mm-hmm.

Mohcine Chraibi 7:38

And, um, we ask from the software too much and, w- the students started complaining and say, "No, I'm gonna write my own script and I don't want to use JJuPedSim."

Wojciech Wegrzynski 7:47

Okay. Okay.

Mohcine Chraibi 7:48

It's faster this way."

Wojciech Wegrzynski 7:49

It's faster, okay. Yeah.

Mohcine Chraibi 7:51

You know? And this was the breaking point where we said, "Okay, if even our student cannot use it, so what's the point?"

Wojciech Wegrzynski 7:57

W- w- which is kind of interesting, you know, because Th-th-that's not your unique experience. I think that's very true for a lot in fire science, where people would just, you know, solve their part to create a model and, then eventually move to other things. There's a very typical story of, of someone becoming a PhD student or a postdoctoral researcher working on a specific part of a specific project, which perhaps is extremely useful or valuable to the community, and perhaps they have energy to s- to, you know, keep and maintain this for two, three years or five years. But then eventually life, catches up and it's to some extent abandoned, not from theee research validity, but from the, technical maintenance perspective. Yeah, exactly. Right? And-

Mohcine Chraibi 8:39

Yeah, I mean, we, we take software... In, in academia, we take software for granted.

Wojciech Wegrzynski 8:43

Yeah.

Mohcine Chraibi 8:43

We think, yeah, software, uh... Everyone needs software. everyone needs to implement the model somehow and do the simulations, right? But then who will take care of this all? And y- we have examples like, uh, MENGAR. Mm-hmm. This was a software, very nice, from America, but now it's abandoned because the guy behind it finished his PhD, and then now he's doing something else.

Wojciech Wegrzynski 9:03

Mm-hmm.

Mohcine Chraibi 9:04

FDSEvac, as you know.

Wojciech Wegrzynski 9:05

Yes.

Mohcine Chraibi 9:05

Same story.

Wojciech Wegrzynski 9:06

Th-that's actually an interest because it's important for the, for this story, right? So- Yes maybe, maybe, uh, not everyone remembers E- Evac, w-what that has been.

Mohcine Chraibi 9:14

Yes. Uh, I mean, FDSEvac, it's, it's, uh, the, the software, it was NIST FDS implemented directly. It implemented the social force model. It was a very nice software. It also interacted with fire, which i- which make it very appealing for the human behavior- Mm community. But at some point, the maintainer, it was a one-man show. The guy had s- as you say, life catch up to him, and then he has to do something else.

Wojciech Wegrzynski 9:38

Mm. So, uh, uh, FDSEvac was, w-was this in version five or removed? Or-

Mohcine Chraibi 9:43

I think in FDS five it was-

Wojciech Wegrzynski 9:45

I, I, I

Mohcine Chraibi 9:45

don't- not anymore there.

Wojciech Wegrzynski 9:47

Uh, maybe in some version of five it's still... I, I, I don't recall exactly when they... But I still remember playing with it. Uh, it was, uh, a little bit rough, but the, the beauty of it- Yes was that you could actually run the evacuation layer on top of, simulation, uh, FDS simulation layer, and you could visualize it in, in Smokeview, right?

Mohcine Chraibi 10:07

Exactly.

Wojciech Wegrzynski 10:07

Yeah. Yeah, yeah. There were people walking. That,

Mohcine Chraibi 10:09

that's because there is the whole ecosystem around FDS, and then it's... You could, like, just do zero simulation. You have your FDS data simulations. You can visualize it. Everything is fire Works. Okay, but then the problem that you always have in software is that you have one maintainer, and without this one maintainer nothing works.

Wojciech Wegrzynski 10:27

Mm-hmm.

Mohcine Chraibi 10:28

Luckily, for FDS, they have, like, four or five guys- Mm doing nothing but that.

Wojciech Wegrzynski 10:33

Yes, yes.

Mohcine Chraibi 10:33

And this is a good, a good, uh, model because, uh, those guys, they are not researchers doing, like, research and then when they have- To understand, yeah some free time they do software.

Wojciech Wegrzynski 10:44

Yeah, yeah.

Mohcine Chraibi 10:44

I mean, they do software mainly.

Wojciech Wegrzynski 10:46

Yeah, yeah, exactly. You know? And if they do research it's to improve the software and-

Mohcine Chraibi 10:49

Exactly But, but they do the smoke view, they do the, the model and the validation and the testing, and this is a lot of work.

Wojciech Wegrzynski 10:56

You know? O- of course, and, uh, yeah, I, I also agree that it's, it's, you know, uh, like in the memes where the, the whole like, uh, puzzle of modern computing stands on, on a little block of some guy- Yeah maintaining a library because he likes it. Yeah, yeah. You know? It's- Yeah, yeah I, I think, uh, we, we on, on philosophical terms like, uh, I, I would say that organizations like, uh, IAFSS could perhaps step in to provide a framework to maintain those, uh, key, you know, elements. I, I know it's not a... Perhaps it's a, "I'm not in the leadership of IAFSS anymore," but, uh, I, I, I, I would say that th- this is perhaps something like, uh, if it's for the community, community should take care o- of those softwares.

Mohcine Chraibi 11:38

Uh- Yeah. It, it would be nice I think. Actually, this is, if the community cannot take care of the software, then the software is not gonna-

Wojciech Wegrzynski 11:44

Yeah

Mohcine Chraibi 11:44

stay for long.

Wojciech Wegrzynski 11:45

Okay. But you know, it's, it's not that the EVAC was the only software for, uh, evacuation. Uh, there, there's like, of course, Pathfinder that's probably has the most use in, in fire, but there are, uh, there, there's Building Exodus from- Yes from Greenwich. There's, Oasis.

Mohcine Chraibi 11:58

There is Simulx.

Wojciech Wegrzynski 11:59

Yeah.

Mohcine Chraibi 12:00

But there are a lot of commercial software of course. So for engineers- Mm uh, I think they can survive. I'm talking now just about the research community.

Wojciech Wegrzynski 12:07

Yeah.

Mohcine Chraibi 12:07

So if you want to, to do your own model, okay.

Why Researchers Need Open Tools

Wojciech Wegrzynski 12:10

Yeah. So what, what are the features of the models that you need as a researcher that perhaps the, commercial packages are challenging for you?

Mohcine Chraibi 12:17

Uh, with respect to evacuation or to- Yes. With- or to fire

Wojciech Wegrzynski 12:20

simulation? Well, with respect to evacuation- Yeah and then may perhaps in ge- ge- general, because I, I think a lot of this will

Mohcine Chraibi 12:24

be gen- I mean, researchers they keep doing new research and developing new ideas, right? Yeah. New algorithms and whatever, and it has to be implemented somehow. I think in the commercial software it's not like out of the box you can just call them and say, "Can you please implement my model?" Or- Mm. or, "Give me access to your code so I can..." It's a bit difficult like this. So for, for engineers, okay, I can just use it.

Wojciech Wegrzynski 12:48

Mm.

Mohcine Chraibi 12:49

Sure. But if you are, if you want to develop your ideas and implement them and test them, validate them, whatever, in your own way, I think you need to have access to the code.

Wojciech Wegrzynski 13:00

H-h-how about the, the, the commercial packages being kind of, uh, black boxes or environments that you sign into and, uh, you don't even- Yeah. Yeah you don't have all the knobs to, to- Exactly to fiddle with.

Mohcine Chraibi 13:10

Yeah. Uh, uh, take an example of software. Uh, uh, let's not say names, but, um... No, it's fine. Pathfinder has one model.

Wojciech Wegrzynski 13:18

Mm.

Mohcine Chraibi 13:18

Yeah. But, um, what if I don't want to use that model, but I want to use, some, some- One

Wojciech Wegrzynski 13:24

that you, one that you're developing

Mohcine Chraibi 13:25

right now. Exactly. What I did or some other version of social force

Wojciech Wegrzynski 13:29

model. Mm-hmm.

Mohcine Chraibi 13:29

There are many versions of it. Maybe I want to know what's the difference of all those version and I want to implement them and then do a paper about comparison of models. I can't do this with Pathfinder.

Wojciech Wegrzynski 13:39

Well, you, you can call Brian and, and try- Yeah, of course to convince him to implement it- But- but it's his goodwill and a lot of friction and time. Exactly. Exactly.

Mohcine Chraibi 13:47

Maybe he doesn't see the monetary value in it.

Wojciech Wegrzynski 13:50

Or it's just not in the current development schedule and- Yeah,

Mohcine Chraibi 13:53

sure. But if you are a client, you are paying money, and you say, "I need a new feature"- Mm sure, that will happen sooner. But, uh, I, I just want to publish my paper because I have a PhD to finish.

Wojciech Wegrzynski 14:03

Yes. Yeah. I mean, the Gr- Green- Greenwich kind of has that with Exodus. They can implement their own stuff in the, in the building Exodus- Yeah,

Mohcine Chraibi 14:09

exactly

Wojciech Wegrzynski 14:09

for a, for a way.

Mohcine Chraibi 14:10

But, but, but that's the same for UPETIM now.

Wojciech Wegrzynski 14:12

Yeah.

Mohcine Chraibi 14:12

Now, so UPETIM, we develop it in such a way, it's not like Pathfinder. Mm. Like you have a, a program, you know, with GUI and everything visualization. No, it's a library.

Wojciech Wegrzynski 14:22

Mm-hmm.

Mohcine Chraibi 14:22

documented, clear API, everyone can use it and build on top of it-

Wojciech Wegrzynski 14:27

Mm

Mohcine Chraibi 14:27

easily. Mm. So this docking in is, is the point. So people now use UPETIM in that way. They, they write their Python code and do something crazy. It doesn't affect UPETIM- but it brings results faster-

Wojciech Wegrzynski 14:40

Mm

Mohcine Chraibi 14:40

in that way to the people without, uh, making any trouble for us as, for as a...

Wojciech Wegrzynski 14:45

So- So you, you release a library, and then person can- Exactly adjust again.

Mohcine Chraibi 14:48

You just import it and say, "Okay, now UPETIM can tell you I can move people from A to B."

Wojciech Wegrzynski 14:55

Mm-hmm.

Mohcine Chraibi 14:55

Tell me how." And then you write, "Okay, now I want you to go there and stand, um, for a while, and then if there is fire, do something else," and whatever. So the creativity comes from the community, and we just deliver the-

Wojciech Wegrzynski 15:08

Framework

Mohcine Chraibi 15:09

exactly.

Wojciech Wegrzynski 15:09

Perfect. Um, let's, let's perhaps talk about the structure of JuPedSim. In your talk, you, you've shown, uh, multiple building blocks that, that are necessary to carry out, a complete, let's say, simulation of, of human in fire, uh, uh, from the perspective of this package. So, so, uh, the starting point I understood is, is, uh, a fire simulation like FDS. But I... You could run it without the smoke. It's just- You

Mohcine Chraibi 15:35

can run it also without smoke.

Wojciech Wegrzynski 15:36

Yeah. Yeah, yeah. But, but I, I think the focus of this community would be- Exactly smoke. It's smoke. So let's, let's discuss

The PyFDS Evac Toolchain Explained

Wojciech Wegrzynski 15:40

this

Mohcine Chraibi 15:40

case. So, so, so what we are talking about, PyFDSEvac.

Wojciech Wegrzynski 15:43

Okay.

Mohcine Chraibi 15:44

Uh, so PyFDS-

Wojciech Wegrzynski 15:45

So fir- fir- so first FDS. FDS, you run your simulation. Okay. So- You don't have to modify it

Mohcine Chraibi 15:48

first you have FDS. You do your simulation of your room however you want. We don't, we don't really pay too much attention what... how you do it, how you model it. Your FDS files, it's belongs to you.

Wojciech Wegrzynski 15:58

Do you have to specify anything special in FDS- No like special device, slice file?

Mohcine Chraibi 16:02

No, no, no. You just do your simulation. You just give us basically just the result.

Wojciech Wegrzynski 16:06

Just the FDS file as

Mohcine Chraibi 16:08

it is. Yeah, exactly.

Wojciech Wegrzynski 16:08

Output.

Mohcine Chraibi 16:09

Exactly. And then because we don't want to deal a lot with FDS and its data, we have FDS Reader.

Wojciech Wegrzynski 16:16

Mm-hmm. FDS Reader. Okay.

Mohcine Chraibi 16:17

Which is, uh, developed by Lukas' group. So FDS Reader takes a lot of work. Uh, how do you deal with FDS? It's like a nice wrapper. It says, "Okay, now give me extension coefficient."

Wojciech Wegrzynski 16:28

Mm-hmm.

Mohcine Chraibi 16:28

With those condition in this place, in this time, whatever.

Wojciech Wegrzynski 16:32

F- f- for, for, uh, context, we're used to, you know, processing the, the FDS files through, through Smokeview, and Smokeview actually does that. Yeah. Like Smokeview is a- Okay reader of FDS files- Yeah, yeah where it can read and, and demonstrate those. Yeah. But for your needs, what you, you don't need a Smokeview graphics. You need a table.

Mohcine Chraibi 16:48

Exactly. We need... Yeah, exa- we need a library, okay-

Wojciech Wegrzynski 16:50

Yeah

Mohcine Chraibi 16:50

that we can call the functions.

Wojciech Wegrzynski 16:52

Yeah.

Mohcine Chraibi 16:52

So without sticking to some GUI or what... So it's more flexible and easier. Mm-hmm. And like we just pip install FDS Reader and say, "Okay, that's what I need." So it's a very convenient wrapper-

Wojciech Wegrzynski 17:03

Mm-hmm

Mohcine Chraibi 17:04

around FDS. And, most of the things that you need from FDS, they are always the same basically.

Wojciech Wegrzynski 17:10

Mm.

Mohcine Chraibi 17:11

You always need some quantities, and then during some time and some positions, and FDS Reader does this for you.

Wojciech Wegrzynski 17:16

Mm-hmm.

Mohcine Chraibi 17:17

So for us, it means also it's already done. We don't need to do this again. We don't need to start from scratch. So this block is very important. Then you have the other block, which is, Okay, what about visibility? Now we have smoke, and we have heat, and we have everything. Uh, can we see the signs or not?"

Wojciech Wegrzynski 17:32

Mm-hmm.

Mohcine Chraibi 17:33

And this is where the other library comes in, FDS Vis Map. Mm-hmm. It's also Python library, um, maintained by Christian so FDS VisMap also tells you things like, things that you would like to know as, uh, as a JuPedSim user.

Wojciech Wegrzynski 17:48

Mm-hmm.

Mohcine Chraibi 17:48

What you want to know. You are here, I want to go there, so how does it look like smoke-wise? Mm-hmm. Can I see the exit or not? on things like that.

Wojciech Wegrzynski 17:58

So, so basically it's a, a wrapper for visibility. Uh, we, we just finished, uh, we're up to the, the visibility- Yes symposium, where for three days we were discussing the inadequacy of that, and, uh, I, I'm planning to have a full podcast episode with Lucas about, what we've learned so far in our projects and, and s- and some broader view o- on that. So for the listeners, we will come back to this topic. The important, uh, right now is that instead of basing on local visibility, you use the ability of the, the V- VisMap library to predict whether s- a pathway is- Exactly you can see through it or not. Yes. Right?

Mohcine Chraibi 18:35

Exactly.

Wojciech Wegrzynski 18:35

So, so first FDS, then wrapper, then VisMap, and what's next?

Mohcine Chraibi 18:39

So now you have all the components you need. So I can- From

Wojciech Wegrzynski 18:42

the

Mohcine Chraibi 18:42

smoking

Wojciech Wegrzynski 18:43

environment

Mohcine Chraibi 18:43

side. Yeah, exactly. Now I have JuPedSim. JuPedSim, it's, it does nothing but it tells me, "Okay, now you have your agents. I do the routing for you." Mm-hmm. "I send them to some places," and whatever. Just basic evacuation.

Wojciech Wegrzynski 18:58

Okay.

Mohcine Chraibi 18:58

You can also tell them to go somewhere and wait. This is JuPedSim part. Mm-hmm. It gives you the routing and the operational level. It moves people from A to B. Now, with those, with all those inputs, so the FDS reader inputs, with the visibility inputs, and with JuPedSim, we can now write our own code- Python, which we will call PyFDSEvac pyFDS EVAC because it's like FDS EVAC- Mm just in Python. And this pyFDS EVAC will tell JuPedSim, "Ah, look, this road is not so visible because FDS VisMap told me so."

Wojciech Wegrzynski 19:32

Ah, okay.

Mohcine Chraibi 19:33

And by the way, the extinction coefficient is that much, so I have to be slow. Okay? And we will implement everything with FDS EVAC implemented and, and even more because now we can do more. And then you can tell JuPedSim on the Python l- layer, This is the road I would like to go."

Wojciech Wegrzynski 19:51

Mm.

Mohcine Chraibi 19:51

And this is how fast I should be." And maybe now I should just have, uh, desired speed equal to zero because, you know, you inhale a lot of smoke.

Wojciech Wegrzynski 20:00

This is the part I didn't fully capture from the talk. I'm really glad that we're doing a follow-up. And now I have to rename the episode because I thought it... We're-- I'm gonna name it JuPedSim, but I should call it pyFDS EVAC.

Mohcine Chraibi 20:12

I think

Wojciech Wegrzynski 20:12

it's a- Because that's the-

Mohcine Chraibi 20:13

a better name, yeah. Yeah. And, and I have the talk online, so you can also have- Fantastic

Wojciech Wegrzynski 20:16

have

Mohcine Chraibi 20:16

it later.

Wojciech Wegrzynski 20:17

But, uh, so, so you could consider JuPedSim as the, like, engine of movement and, and, wayfinding and, and uh, pyFDS EVAC the tool that, uh, kind of Narrows it down because of the fire situation,

Mohcine Chraibi 20:31

right? Exactly. So, so, so JuPedSim and FDS are basically in this pipeline the same.

Wojciech Wegrzynski 20:36

Mm-hmm.

Mohcine Chraibi 20:36

Just on different ends.

Wojciech Wegrzynski 20:38

Okay.

Mohcine Chraibi 20:39

Okay? And, and FDS VisMap and, uh, GPS reader, they are just helper-

Wojciech Wegrzynski 20:45

Mm-hmm

Mohcine Chraibi 20:45

libraries to deal with FDS because we don't want to deal directly. But FDS and JuPedSim, they do the same. Just this one is for fire, and this one is for the pedestrian.

Wojciech Wegrzynski 20:54

But they're both transmitting one directionally to PyFDS Ever, right? So- Yeah, of course,

Mohcine Chraibi 21:00

yeah.

Wojciech Wegrzynski 21:01

Okay.

Mohcine Chraibi 21:02

That's, of course, because, uh, we don't influence fire anymore. Fire is done, the simulation.

Wojciech Wegrzynski 21:07

I know, but, uh, I mean, technically with this capability, one thing that you could do is the state of the doors because you know if someone is currently at the door or not. Yeah,

Mohcine Chraibi 21:17

yeah, yeah.

Wojciech Wegrzynski 21:17

Which could technically influence the flow field in FDS. That's something that, uh-

Mohcine Chraibi 21:23

That, that could come, but then we will need to have a talk with the FDS guys.

Wojciech Wegrzynski 21:29

You can, uh, you can definitely, uh, you know, have, uh, walls that disappear and appear. Yeah. Uh, but, but the reason for that is that So out- outside of the, of the room before we started, uh, we, we had this chat when you asked me like, "Do, do you think people really need it?" Like, because engineers like to have a simple calculation and be done- Yes with it, you know. For example, for the UK market and smoke control in, buildings, there is this specific prescribed sequence of stuff that you need to include in your design fire scenario. I have a fire in my room. I open the doors for 60- Yeah seconds. I close the doors for 60 seconds, and, uh, then this corridor cleans for this amount of minutes, which again is like kind of capturing the human impact on the environment around them in a prescribed scenario. Prescribing which doors were open, which doors were closed. If you have this, uh- Yeah you know, in the simulation, it makes a big difference, uh, because-

Mohcine Chraibi 22:26

Yeah, that, that's what you say. I mean, our motivation, it was not I mean, we are scientists, and we s- we do sometimes science even if no one needs it.

Wojciech Wegrzynski 22:35

Mm.

Mohcine Chraibi 22:35

And maybe later someone will pick on it and say, "Okay, this is useful."

Wojciech Wegrzynski 22:39

I, I'll give you another example. We, we are designing a, a, a, a tramway system and an interim stage of that s- system is that there's a tunnel leading to a station which has, um, one side closed.

Mohcine Chraibi 22:51

Okay.

Wojciech Wegrzynski 22:52

In next future, there's gonna be another tunnel, another station- Mm and, and way out. But for now it's a, a single directional tunnel connecting to a station. Okay.

Mohcine Chraibi 22:59

Yeah.

Wojciech Wegrzynski 23:00

And because we have extraction in the tunnel by jet fans, the air comes into the station and basically we, we create the movement of air to throw the smoke out. Now this creates also an, under pressure in the station, like negative pressure and, There's also a limited amount of air we can have to that station by mechanical means. We don't to build too many shafts et cetera. But, uh, w- we do simulations for this state, but in reality, as soon as people start escaping this, uh, station, they open the doors and they create like, you know, new pathways of air. Yes. So the resulting, you know, under pressure is much less severe- Mm and the amount of air we get there is, is much higher and there is like the performance of the system becomes so much better- Yes because I have so much more air available. Of course. But, uh, the-- I currently do not have a-

Mohcine Chraibi 23:47

Yeah

Wojciech Wegrzynski 23:47

a tool that could robustly allow me- Yeah to show this effect because it is the, the human impact on the fire scene- Mm that I need to, introduce.

Mohcine Chraibi 23:58

But this is, this is definitely interesting, Wojciech, but for now we, we will be just unidirectional.

Wojciech Wegrzynski 24:03

I know. I, I- Again, I can go and bother Brian about it

Mohcine Chraibi 24:07

Yes, you should do that.

Wojciech Wegrzynski 24:09

Definitely And, uh, and uh, and uh, it's gonna take them a lot of time a- and, and thinking, and they can put it on a map. Or perhaps there's a, you know, PhD student that's interested in a topic like that- Mm that can take your library and start working on it immediately.

Mohcine Chraibi 24:23

Yeah. I mean, it's, it's very extensible and, and things like that can be done easily. I mean, I'm, I'm not sure in the other direction when you go to FDS what you can do because I don't do FDS.

Wojciech Wegrzynski 24:33

Mm-hmm.

Mohcine Chraibi 24:34

But, uh, in your Petium direction, uh, sure. It's, uh-

Wojciech Wegrzynski 24:37

Okay. So, so from the pra- practical perspective, because it's, it's five packages, uh, there, there's a lot of Python codes, there's likely a lot of terminal. Like, what, what kind of skill set do you, do you need

Installation Options And Web Interface

Wojciech Wegrzynski 24:50

to, to run this?

Mohcine Chraibi 24:50

Uh, none. Um, If you, if you know how to work with Python-

Wojciech Wegrzynski 24:55

Mm-hmm

Mohcine Chraibi 24:55

then you just pip install pyfdsvac

Wojciech Wegrzynski 24:58

And that's it

Mohcine Chraibi 24:58

And then it will install everything for you-

Wojciech Wegrzynski 25:00

Okay

Mohcine Chraibi 25:00

apart from FDS, of course.

Wojciech Wegrzynski 25:02

Okay

Mohcine Chraibi 25:02

Just the Python library is around. This is everything will be installed and set up. You just run pyfdsvac and it knows everything is already installed. You have to point it, of course, to some directory where you have your FDS data- Mm-hmm and say, "Those are my FDS data. Please do your magic now." Alternatively, if you don't know how to do Python, we have this web-based interface. This is a bit experimental now because we don't know yet if people need it or not. We don't need it in our group because- Mm we do the Python thing.

Wojciech Wegrzynski 25:32

You prefer the Python, yeah. Mm-hmm.

Mohcine Chraibi 25:33

We are faster this way, but I can imagine that people just want to know, "Okay, I will click here. This is my FDS file."

Wojciech Wegrzynski 25:39

I, I, I think, uh, you know, for, uh, again, you, you're researching it. Uh, you, you are building it as is, and you, you let it, uh, for people to use. I, I think for engineers there will be a lot of friction.

Mohcine Chraibi 25:52

Yeah,

Wojciech Wegrzynski 25:53

true. I mean, FDS also to a sense has, you know- Yes this, uh, this... It's kind of this hacking when you write the FDS code. But in reality, in projects, like most of the people would use some sort of pre-processor like- Yeah like PyroSim to, to, to just get... It's just easier that way and, and perhaps more natural. Also, you know, we're civil engineers. Like you're very used to working in AutoCAD or- Yeah or any other like

Mohcine Chraibi 26:20

device. No, definitely.

Wojciech Wegrzynski 26:21

Yeah. So-

Mohcine Chraibi 26:21

I, I think so. But this is also why, I mean, there are, there are those tools that use FDS as an engine to shield people from all those details.

Wojciech Wegrzynski 26:29

A, a, a question that, that follows. the, the building itself, the walls, the doors, the paths, the exits Is it directly transferable from FDS file or you-

Mohcine Chraibi 26:40

We, we extract them from the FDS file.

Wojciech Wegrzynski 26:42

Okay.

Mohcine Chraibi 26:42

So to us to- so that we can have only one- In- source of truth.

Wojciech Wegrzynski 26:47

Okay.

Mohcine Chraibi 26:48

We don't want to have... I mean, the simul- the FDS simulation is important, and it's done with FDS file. Yeah. And FDS file has some geometry. Yeah. So we will try to get it.

Wojciech Wegrzynski 26:57

Okay. But there will be things that will be difficult, like stairs.

Mohcine Chraibi 27:00

Like- Yeah, sure. I mean, uh, JuPedSim right now-

Wojciech Wegrzynski 27:04

Mm-hmm

Mohcine Chraibi 27:04

it's just 2D.

Wojciech Wegrzynski 27:05

Okay.

Mohcine Chraibi 27:06

And we are working on JuPedSim version 2.0 will be released in December- Okay 26, with stairs and- With stairs two and a half D. And, so for now it's just 2D everything.

Wojciech Wegrzynski 27:16

Mm-hmm.

Mohcine Chraibi 27:17

Uh, but of course we are moving to the building level with stairs and also FDS VisMap I saw. Did you see- Yes,

Wojciech Wegrzynski 27:25

yes

Mohcine Chraibi 27:25

Christian's talk?

Wojciech Wegrzynski 27:26

Yeah, he's

Mohcine Chraibi 27:26

trying- He's also trying to do 3D- 3D,

Wojciech Wegrzynski 27:28

yeah

Mohcine Chraibi 27:28

visibility maps. So it's all merging and came, uh- Mm-hmm converging to the same

Wojciech Wegrzynski 27:34

important because of course like, you know, vertical, uh, escape routes- Yeah, sure are, are absolutely- Of course, of course absolutely critical to, to, to be integrated in the model if we want it to succeed. And n-now, I mean, wh- when you start having this complex geometry, it's gonna be difficult to... Because, uh, f-for example, me as, you know, a modeler, I don't care if a thing is escalator. Is it moving stairs? Yeah. Is it just stationary? For me it's just an obstacle. Exactly. Yeah, sure. Uh, and thing for you it's probably-

Mohcine Chraibi 28:00

No, I mean, if you, if you do proper modeling in FDS file-

Wojciech Wegrzynski 28:04

Yeah

Mohcine Chraibi 28:04

of your geometry-

Wojciech Wegrzynski 28:05

Mm-hmm

Mohcine Chraibi 28:05

it should be not an issue for JuPedSim. But if you have there some errors and things are not connected well and whatever, then of course.

Wojciech Wegrzynski 28:12

But for, for now you do not, have any kind of BIM import or, or, or integration with-

Mohcine Chraibi 28:17

So for JuPedSim alone- Yes you can have IFC files imported.

Wojciech Wegrzynski 28:21

Okay.

Mohcine Chraibi 28:22

But, uh, I mean, JuPedSim as- JuPedSim as- as a thing. And maybe th-this is something I didn't tell you, but JuPedSim has also an own web-based interface-

Wojciech Wegrzynski 28:32

Okay

Mohcine Chraibi 28:32

that we built I mean, as a library.

Wojciech Wegrzynski 28:35

Mm-hmm.

Mohcine Chraibi 28:35

It's a library. But some people they said, "I don't want to deal with Python." We had this already.

Wojciech Wegrzynski 28:39

Okay.

Mohcine Chraibi 28:40

So we built this web-based, uh, interface. I can give you the link later.

Wojciech Wegrzynski 28:43

Okay.

Mohcine Chraibi 28:43

It's called app.JuPedSim.org. Mm-hmm. So people, they can click and move and draw and-

Wojciech Wegrzynski 28:48

Okay. So, so this is a version without fire, uh-

Mohcine Chraibi 28:51

Without fire. And we are planning maybe to Interact also to integrate PyFDSEvac EVAC in it.

Wojciech Wegrzynski 28:57

Mm-hmm.

Mohcine Chraibi 28:58

the only problem I have right now is really before we start doing this work, we are not sure if people want it or not.

Wojciech Wegrzynski 29:05

Mm-hmm.

Mohcine Chraibi 29:05

So if the community, somehow it happens that the community kills us with emails and say, "Ah, we need this all," then we will do.

Wojciech Wegrzynski 29:12

Mm-hmm.

Mohcine Chraibi 29:13

But for now, I mean-

Wojciech Wegrzynski 29:14

You, you go directions that is for yourself,

Mohcine Chraibi 29:17

yeah? Exactly, yeah. We, we... L- it's like FDS also. FDS doesn't- Yeah have any web-based stuff. It's just FDS, right?

Wojciech Wegrzynski 29:23

Right. All

Mohcine Chraibi 29:23

right. And people use it and build the ecosystem. It's, it's organic, but-

Wojciech Wegrzynski 29:27

No, of course, like you're also not, uh, not selling licenses- No, we don't uh, which would give you, uh, uh, immediate feedback. If it sells well, it's in a good direction, uh, and, and learn the pain points of, of the customers. Uh, it's a different relationship. Uh- Yeah but also one that, uh, of the reasons why, why you, you choose to build it is on... Can you tell me about, uh, more of the, about the engine? Uh, uh, how does the movement work? How does the wayfinding

Movement Models And Switching Behaviors

Wojciech Wegrzynski 29:52

work?

Mohcine Chraibi 29:52

Yes.

Wojciech Wegrzynski 29:52

I- I... Like did you, did you had to come up with some own ideas for, for that? Or, or is it well known and you just apply it?

Mohcine Chraibi 30:00

No. Uh, I mean, there is, uh, there is in, in, in our community this model. It's, uh, called, uh, it's a Hogendoorn model. Mm. It's like you have strategic level, tactical level, operational level.

Wojciech Wegrzynski 30:11

Mm-hmm.

Mohcine Chraibi 30:11

In the operational level, you have the social force model, for example.

Wojciech Wegrzynski 30:14

Mm-hmm.

Mohcine Chraibi 30:14

That's just you move from A to B, and at the tactical level, you, you do the routing and you go somehow through the exits. And then the strategic level, you ha- you do something more complicated like, uh, which doors should I take into consideration? Mm-hmm. What do I want actually? Do I want to evacuate or whatever? And in uPSIM, we implement all those three levels.

Wojciech Wegrzynski 30:33

Mm-hmm.

Mohcine Chraibi 30:34

On the operational level, we have, I'm not sure, five different models.

Wojciech Wegrzynski 30:40

Okay.

Mohcine Chraibi 30:40

Like social force model, classic.

Wojciech Wegrzynski 30:42

W- what does it do? It keeps people apart from each other?

Mohcine Chraibi 30:45

Yeah. This... The, the, the operational model is the model that tells you, "Now I have to go, and now I have to avoid the others." Okay. "And I have to avoid the walls."

Wojciech Wegrzynski 30:53

Okay.

Mohcine Chraibi 30:54

So on this level. So this is just the basic, like social force model. It say- says, "Okay, there is a door there. Now I have to move." Right?

Wojciech Wegrzynski 31:00

All right.

Mohcine Chraibi 31:01

Stepwise, and then there are some bunch of people, I have to avoid them.

Wojciech Wegrzynski 31:04

Just- You don't bump into them, okay

Mohcine Chraibi 31:06

don't bump into the walls too.

Wojciech Wegrzynski 31:07

Mm-hmm.

Mohcine Chraibi 31:08

This kind of logic. And this kind of logic is modeled or is expressed in terms of the social force model-

Wojciech Wegrzynski 31:15

Mm-hmm

Mohcine Chraibi 31:15

which is very known, and everybody likes to use it. So we have it too. But we also have other models like velocity-based model or- Uh, warp driver model, whatever models we find in the, in the literature that are interesting and good- Mm-hmm we think. interesting maybe. So we have five. That means the user, can choose. I- now, for this case, I can use the social force model, but maybe for the other case I can-- I have to use something else.

Wojciech Wegrzynski 31:42

Mm-hmm.

Mohcine Chraibi 31:43

Now, the most important part is coming, and this is a feature that we are working on. It's already... It's not yet released, but we plan to release in two point zero. What if you are in this building and you say, "Okay, for this room, for example, social force model is the best model. But then on the stair, I have to, to change to another stair model."

Wojciech Wegrzynski 32:04

Okay.

Mohcine Chraibi 32:05

And then there is a corner, and there is a corner model. You know? So what do I do? I can now switch between those models.

Wojciech Wegrzynski 32:11

Mm-hmm.

Mohcine Chraibi 32:11

Like the agents can... has this awareness and say, "Okay, now I'm using social force model. And after a while I will change-

Wojciech Wegrzynski 32:18

Mm-hmm

Mohcine Chraibi 32:19

completely my behavior, so to say." Uh, this is, this is very cool. You, you can't have this with mo- with the software that have only one model, of course. Mm-hmm. Obviously. But, for us at the beginning, the motivation was only we want to have the flexibility to compare, and we don't want to stick with one model only. But now we can switch between them even.

Wojciech Wegrzynski 32:38

in a bigger picture of, of development, w- was it always the goal to simulate the, the whole buildings, or is it more like a link to the experimental part at the Juelich where they are studying impact of- Yeah for example, narrow wings or a simpler layouts- Yeah but not entire buildings?

Mohcine Chraibi 32:54

Uh, there are two phases. At, at the beginning, we... of course, we are like, uh, we do the, the bottleneck experiment, the corner experiment-

Wojciech Wegrzynski 33:01

Mm-hmm

Mohcine Chraibi 33:01

the bi-directional flow ex- Those are simple setups. And there is the fundamental diagram then, and then you say, "Okay, can I reproduce this with my model?" So our geometries, they are very basic at the beginning. And the, the difficulty is, are the models good or not? Are they validated or not, and why not? and then later, after we redesigned JuPedSim as a library, we had a project, it's called Chroma Pool. And in this project, the, the main focus was in open air events, like peak and open air, which is two-dimensional. So that was the main focus of JuPedSim. Mm-hmm. So we... It has to be fast Enough. It has to be able to simulate a lot of people, and it's just 2D. So we didn't need the complexities of stairs and whatever. But the project is over. Now we move to the third stage, which is now we would like also to do buildings because of PyFDS Evac. So I mean, it makes no sense to have PyFDS Evac just in 2D.

Wojciech Wegrzynski 34:00

In f- for one room.

Mohcine Chraibi 34:01

As you, you just said, yeah.

Wojciech Wegrzynski 34:02

Mm-hmm.

Mohcine Chraibi 34:02

You, you need to simulate the building, otherwise,

Wojciech Wegrzynski 34:05

Tell me about the, the pathfinding because that's also an interesting aspect- Yeah of every evacuation structure. Yeah, sure. So how, how does the agent know that you have to go from here to the safety?

Pathfinding From Graphs To Flow Fields

Mohcine Chraibi 34:15

Yes. So, our algorithm is very similar to pathfinders. So we have the building, and then we, code the building in form of graph And then you have the nodes and the edges, and then you have to find the shortest path along the graph-

Wojciech Wegrzynski 34:30

Mm-hmm

Mohcine Chraibi 34:30

with Dijkstra or of A-star. Uh, it's called A-star algorithm, but it's basically just finding the shortest path on the graph. That's it, first. And then now, as I said, we are moving to two and a half, two and a half D, so that means we need different algorithm, and we are implementing flow fields, uh, which means basically, um... So it's like you can imagine you have the door, and from the door you have a, a source of light or a source of- Mm-hmm water or whatever, and then you can flood the room.

Wojciech Wegrzynski 35:01

Okay.

Mohcine Chraibi 35:02

So the water, it doesn't move randomly. It moves in waves. In

Wojciech Wegrzynski 35:05

waves, okay, okay.

Mohcine Chraibi 35:06

So those waves we can, like... I mean, we can keep them as file or in the memory, and then the a- agents, they just move along and then find their way. Perpendicular

Wojciech Wegrzynski 35:15

to the waves.

Mohcine Chraibi 35:15

Yes, exactly.

Wojciech Wegrzynski 35:16

Okay.

Mohcine Chraibi 35:17

And if you do this for the whole building, then the pedestrians find automatically their way.

Wojciech Wegrzynski 35:21

Mm-hmm.

Mohcine Chraibi 35:22

Which is, you can say it's the same like the graphs, but why do we do this? But this one has the benefit that it's allows you to have smoother paths, and you can also manipulate the wave with some dynamical forces.

Wojciech Wegrzynski 35:35

I, I may be, I may be wrong. Please send complaints, but I think Exodus works like this. I, I think they, they-

Mohcine Chraibi 35:40

Maybe, yeah. And Simulex also.

Wojciech Wegrzynski 35:42

Yeah, yeah.

Mohcine Chraibi 35:42

Simulex have the same. So it's, it's just a standards algorithm, actually. Mm-hmm. We didn't implement it at the time because for the open air scenarios-

Wojciech Wegrzynski 35:51

Mm-hmm

Mohcine Chraibi 35:51

it was too expensive.

Wojciech Wegrzynski 35:53

Mm-hmm. Oh, right. Okay.

Mohcine Chraibi 35:54

So huge memory. But if you have the huge memory and you are done with it, then it's just, uh, lookup tables. But, we thought, "Nah, we do the graph thing. It's faster and it's better." but now I think for the buildings we will, uh, move to pathway.

Wojciech Wegrzynski 36:09

With, with graphs you also have shown a very interesting, uh, part of your talk, uh, which was the, the, the wayfinding for non-familiar people- Yes where, uh, basically a person doesn't start with a graph. They have to discover. Can you, can you explain that concept? Yes. I found it

Familiarity And Cognitive Map Wayfinding

Wojciech Wegrzynski 36:24

very interesting.

Mohcine Chraibi 36:24

Well, it's, it's based on, on a, a concept called cognitive maps.

Wojciech Wegrzynski 36:27

Mm-hmm.

Mohcine Chraibi 36:28

Um, it's... Biologists did that already. They know in, in the mind of- Mm-hmm people how it work. But with... There are some papers that we, uh, that tried to implement it for the pedestrian dynamics. Mm. But basically what it says is, you have this graph of the building, and of course if you live in this building or you, you work in this building, then you know you have access to the whole graph. You know everything, every single corner of this building. But-

Wojciech Wegrzynski 36:50

So you can pick the shortest path from-

Mohcine Chraibi 36:52

Exactly

Wojciech Wegrzynski 36:52

you to the exit, Dijkstra- Exactly and you're done.

Mohcine Chraibi 36:54

So it's like global shortest path.

Wojciech Wegrzynski 36:56

Mm-hmm.

Mohcine Chraibi 36:57

Because you know everything

Wojciech Wegrzynski 36:58

Because you know everything

Mohcine Chraibi 36:58

But you don't know always everything, especially if you are just visiting.

Wojciech Wegrzynski 37:02

Mm-hmm.

Mohcine Chraibi 37:03

So this familiarity parameter, so one familiarity equal to one would mean you know everything, but familiarity equal to zero would mean you know nothing.

Wojciech Wegrzynski 37:11

So you only see the room you're in, and you have no idea what's behind the door.

Mohcine Chraibi 37:14

Exactly. It's like, uh, I, I mean, you come in this room blindfolded, and then we tell you, "Now you can leave."

Wojciech Wegrzynski 37:22

Mm-hmm.

Mohcine Chraibi 37:22

Now, we did this experiment, by the way, once in Juelich. Mm. Like really, we brought people from Wuppertal-

Wojciech Wegrzynski 37:27

Mm

Mohcine Chraibi 37:27

and they entered the building from the back door. Okay. And we said, "Now please leave the building without turning back."

Wojciech Wegrzynski 37:34

Ah, without turning back. Okay.

Mohcine Chraibi 37:35

Yes. And we removed the signs.

Wojciech Wegrzynski 37:37

You removed, okay.

Mohcine Chraibi 37:38

We rem- uh, we put stuff on the signs so people don't know. And then, yeah, and then, uh, we recorded. They had like, uh, eye trackers, and then we recorded how they find their way. Some people found a way, some took more some time, some took less So that's basically the familiarity.

Wojciech Wegrzynski 37:54

Mm-hmm.

Mohcine Chraibi 37:54

So they had like familiarity zero-

Wojciech Wegrzynski 37:56

Okay

Mohcine Chraibi 37:57

actually. And then- So

Wojciech Wegrzynski 37:57

they had to discover and see where

Mohcine Chraibi 37:59

they get. Exactly. Yeah, yeah. But then this discovery, it means also... So first you have maybe one edge in your graph. Mm-hmm. But then you discover something more, and then you have two edges, and then three. So your graph can grow.

Wojciech Wegrzynski 38:12

Mm-hmm. And how, how does the, the fire environment affect the graph?

Mohcine Chraibi 38:15

So the fire, it's, it's... So two things affect the graph. Visibility, so if you see the, the exit or not-

Wojciech Wegrzynski 38:21

Mm-hmm

Mohcine Chraibi 38:21

of course. And then, to see the exit or the visibility of the exit depends on the angle, if it's, if you are looking at it or not.

Wojciech Wegrzynski 38:29

Mm-hmm.

Mohcine Chraibi 38:29

And it also depends on the smoke between you and, and the exit. So, so there are exit, they are there nearby, but they are not visible.

Wojciech Wegrzynski 38:36

So- So the graph excludes them, and it doesn't have that leg.

Mohcine Chraibi 38:39

Exactly. So, so the, the, the, the non-visible exits, they don't, they don't enter your graph at all.

Wojciech Wegrzynski 38:46

All right.

Mohcine Chraibi 38:47

Yeah. So because you don't see them. Even if they are nearby, but still you didn't see them, so that means your graph is still lacking

How Smoke Slows And Disables Agents

Mohcine Chraibi 38:54

in this.

Wojciech Wegrzynski 38:54

Uh, and in terms of, uh, agent movement, what aspects of the fire environment do you account for?

Mohcine Chraibi 39:01

So apart from now the visibility, we have the speed.

Wojciech Wegrzynski 39:05

Okay.

Mohcine Chraibi 39:06

So that means people, they get slower, uh, with time, depending on the soot- Mm-hmm uh, and the exertion coefficient. And also, at some point, maybe if ED it's a bit equal to one or higher than one, then maybe people they will just stand forever, meaning-

Wojciech Wegrzynski 39:22

Like this agent is lost and-

Mohcine Chraibi 39:23

disabled. Yeah. So they don't, they cannot move anymore.

Wojciech Wegrzynski 39:25

Mm-hmm.

Mohcine Chraibi 39:26

So they, they become also obstacles for others.

Wojciech Wegrzynski 39:28

Mm-hmm.

Mohcine Chraibi 39:29

So that's actually, that's it. So how, what do I know about the room, and, how do I go there? Even if I know now two doors, I have to choose between them.

Wojciech Wegrzynski 39:40

Mm.

Mohcine Chraibi 39:40

And then it's normally it's based on the distance, so you choose the nearest one. Mm. But now maybe you choose not only the nearest one, but also the less smoky one. And then you have the speed

Wojciech Wegrzynski 39:50

D- d- do you also allow for, for things like, social links between people? Like, uh, this person needs to find this person to evacuate?

Mohcine Chraibi 39:58

No. No, no. No, not yet. Right.

Wojciech Wegrzynski 39:59

And there are any, like, assisted evacuation? No, not yet either? No,

Mohcine Chraibi 40:02

not yet. No,

Wojciech Wegrzynski 40:03

no. Uh, what are your thoughts about, uh, decision-making introduction to such models? Like, d- do you think it's, uh, it's something that, that, that you would like to work with? What, what's your s- like, as a researcher in this field, what do you think about it? I

Mohcine Chraibi 40:16

mean, I, I would like to do. Mm-hmm. Yes. But my problem right now is I don't know exactly enough data about this process.

Wojciech Wegrzynski 40:24

Mm-hmm.

Mohcine Chraibi 40:24

So I'm, I'm... I can do some fantasy models and then I think, "Yeah, okay, I think it should be this way." But I feel a bit uncomfortable because I don't know much work with... So this is what we also discussed. Yes, yes. We found out that this is one of the missing gaps. If we could, like, just do s- experiment in just white smoke and then evaluate this decision-making somehow that we can get some empirics, then I think it would make sense to go further in this topic.

Wojciech Wegrzynski 40:53

F-f-for me it's a, it's a challenging thing, you know. Uh, from what you've described right now, it's, it's, uh, pathfinding based on, on the conditions in the building. Like, uh, you either discover the path as you go or, or you know the path and then eventually you modify if you reach, uh, uh, a situation where you don't see. That, that, that's, that's reasonable to me, very reasonable. Uh, the, the fundamental diagrams for speed, I guess, is the higher density- Yes slower speed, uh, et cetera, is also imple- Yes must be implemented. When, when, when we start to do some, like, more advanced things like d-decision making or something, I, I really struggle because one, we have now insane variability- Yes in the, in the- Yeah, yeah inputs, you know? And randomness that requires perhaps thousands if not hundreds of thousands- Exactly of scenarios.

Mohcine Chraibi 41:41

Yes, yes.

Wojciech Wegrzynski 41:42

And, and we all... Like, that, that, that goes back to where, where we started before we st- we pressed record. Like, engineers would like to do one simulation and be done with it, you know? So it does-- I also, like, am very worried about people, taking a very advanced model and say, "Okay. Uh, so this predicts evacuation in this building, and here is how- Yes, yes it's gonna look like."

Mohcine Chraibi 42:01

I'm with you. I'm with you on this. This is definitely... I mean, our job as scientists is to develop s- things, and it, it can be complicated or whatever, or advanced. And then the, the people who, who need to apply them, they have to, to see and to filter out. But having said that on the other hand, you have also this, um, nightclub s-station accident.

Wojciech Wegrzynski 42:23

Yes.

Mohcine Chraibi 42:23

Yes? In Rhode Island? What- Yes when was it? 2003

Wojciech Wegrzynski 42:28

Yes.

Mohcine Chraibi 42:28

or something like that? but, um, I mean, Rita, she studied that later- Mm-hmm after, and then she d- she studied where people were dying, and then she did, uh, she did her report about it. She has a nice paper published. And then what we found out is that familiarity plays a role-

Wojciech Wegrzynski 42:46

Mm-hmm

Mohcine Chraibi 42:46

in this accident. Played one of... one, one role of... So one of the factors was familiarity, so to say. And if my model or my simulation, like NIST did- cannot do familiarity and everyone knows everything, you know, then it's not the situation we have. Mm-hmm. So whatever they simulate and they do, I think it's still lacking. And it's not like, "Oh, this is just a fancy new model that's a bit more complicated." But I think it's necessary in this case to differentiate people based on their familiarity, and then do a sensitivity analysis and say, "Okay. Now let's say 20, 80 or whatever, and then do it 100 times." But then you have a knob to play with.

Wojciech Wegrzynski 43:29

Mm-hmm.

Mohcine Chraibi 43:29

Better than to say everyone has familiarity equal to one.

Wojciech Wegrzynski 43:32

And I guess the same case could be made for, for behavioral decision models, et cetera

Mohcine Chraibi 43:37

Yes

Wojciech Wegrzynski 43:37

uh, different theories that, that are existing in the world of evacuation- Definitely,

Mohcine Chraibi 43:42

yeah

Wojciech Wegrzynski 43:42

which could be, uh, eventually introduced into the model. But with every kind of tool like that, you also need to do 10 more simulations or 10 times more simulations- Exactly, yes to, to see the-

Mohcine Chraibi 43:51

But, but it gives you... maybe it gives you some new insight. Mm-hmm. Maybe it gives you something more extra that you didn't have before.

Wojciech Wegrzynski 43:58

Mm.

Mohcine Chraibi 43:58

Maybe not.

Real Projects, Scale, Dynamic Signage

Wojciech Wegrzynski 43:59

but- On another note, how much do you worry about computational efficiency for buildings? You said open, for op- outdoor you need a, a fast software. How, how much, uh, is the computational efficiency for you?

Mohcine Chraibi 44:10

I mean, uh, compared to FDS, it's nothing.

Wojciech Wegrzynski 44:13

Of course.

Mohcine Chraibi 44:15

Yeah, still, I mean, it's not like you do a simulation for weeks or, or months. I hear stories- Yeah from, from Lucas talking about FDS. Uh, you have to run it on a supercomputer for a month or so, and then I think, "Whoa, that's You are a very patient man."

Wojciech Wegrzynski 44:29

No, you just fire and forget. You start doing other things, and then- Exactly eventually you come back after a month to rediscover your

Mohcine Chraibi 44:34

CFD. Yeah, or do your, your fire simulation and go have vacation and then come back.

Wojciech Wegrzynski 44:37

Exactly.

Mohcine Chraibi 44:38

And then find out that you messed up something.

Wojciech Wegrzynski 44:40

You didn't start the fire.

Mohcine Chraibi 44:41

Exactly, yeah.

Wojciech Wegrzynski 44:42

That, that

Mohcine Chraibi 44:42

happened to me

Wojciech Wegrzynski 44:43

once.

Mohcine Chraibi 44:43

Whatever. But, but for the pedestrians, I think, um, it's not... we don't have this problem. It still can, like-

Wojciech Wegrzynski 44:48

Mm-hmm

Mohcine Chraibi 44:49

it depends. If you have, like, 50,000 people, maybe it takes you a couple of hours. And that's it. And the, the, the... it gets more demanding if you say, "Now I need to do 100 simulations of-" Uh, and the variations and the whatever, then it ca- takes more than a couple of hours, which then it goes beyond

Wojciech Wegrzynski 45:08

O- o- one potential direction for solvers like this that I, I, I see is real-time optimizing the, the evacuation processes in buildings with, uh, through, through the use of dynamic signage and, and kind of information feedback loops with the building users. Of course, this is not something you would do for a single-family household. But if you have a massive, uh, skyscraper or, uh, a- actually, skyscrapers are a really great, um, example because if you have multiple staircases in a skyscraper, you may end up in a situation where one is, uh, abused by people, so it's full- Mm with people and one could be perhaps empty. Yes. And actually, in, in, modern, uh, simulation softwares, you can end up with that because, for example, a pathway to one doors could be a little bit shorter than to another door- Yeah, yeah and most of your floor will empty to this one staircase- Yeah even if in reality- Of course, yeah it could be different, which leads to traffic jams, et cetera. This

Mohcine Chraibi 46:03

is, so this is why you mentioned dynamic, uh, signs.

Wojciech Wegrzynski 46:06

Because you can then- Yeah, yeah, yeah. Okay you could perhaps guide, but, uh, to guide them, you have to have a model that this... Well, you can have dynamic signage that just guides you away from the fire, but when you have, uh, an evacuation of an entire skyscraper, you may have fire on one floor- Mm but you're evacuating 50 other floors.

Mohcine Chraibi 46:22

Yes.

Wojciech Wegrzynski 46:22

So, so you may want, just want to optimize the process. Oh,

Mohcine Chraibi 46:24

and then maybe you, you could end up in a situation where there is no clear solution to it.

Wojciech Wegrzynski 46:30

Yeah.

Mohcine Chraibi 46:31

Like, I mean, uh, go away from the dense, uh, or the clogged exit-

Wojciech Wegrzynski 46:35

Mm-hmm

Mohcine Chraibi 46:36

or go away from fire. Maybe you can do, you can't do both at the same time. Yeah. So this is not a decision I would like to take in your position.

Wojciech Wegrzynski 46:43

it, it has, uh, real-world consequences because you can shorten the evacuation time- Yeah by a significant, uh, uh, time if you, if you do it correctly. Of course. So, so it's mo- more, more like, uh, you know, sa- saving lives not, uh- Yeah, sure not, not just, uh, optimization. But, but I, I, I feel there- With good building model integration, digital twins of buildings and, uh-

Mohcine Chraibi 47:04

I- it could be, I mean, it could be also a good application already if you do the simulation now just with PyFDS Evac as is.

Wojciech Wegrzynski 47:11

Okay.

Mohcine Chraibi 47:11

And you do different scenarios and then you end up with this exit here tends up to be more filled with people-

Wojciech Wegrzynski 47:19

Mm

Mohcine Chraibi 47:20

because of the shortest path or whatever.

Wojciech Wegrzynski 47:22

Mm.

Mohcine Chraibi 47:22

So you could like build up already some knowledge a priori before something- Mm even happens, you know? And then based on that you, optimize your sign models.

Wojciech Wegrzynski 47:32

Another thing that that's, that's kind of, uh, interesting to me and, and, uh, I, I think in your bio I've read you may have some experience with that, which is transportation models. Uh, we do a lot of railway stations- Yes a lot of tramway stations, metro stations, you know? And, uh, you know, my second zero when I press play-

Mohcine Chraibi 47:50

Mm-hmm

Wojciech Wegrzynski 47:50

on my evacuation simulation in reality is a quite a dynamic, uh, situation because it takes a good while for the entire network around me to stop. So I may have a fire starting in my- Yes station, but it doesn't gonna magically disappear a train- Yeah, yeah that's about to arrive, you know? Yes. And, and those dynamic s- scenarios with people coming in, uh, with different modes of, motion-

Mohcine Chraibi 48:14

Yes

Wojciech Wegrzynski 48:14

is something I, I don't think I can capture right now with- Yeah, yeah with my software.

Mohcine Chraibi 48:17

It's, it's difficult. Even this, um... There was a talk this morning, this, um, guy from Denmark. Yes. He said that there is even a fire, but still Uh, they don't... The, the agent, they don't start now, they start still later or whatever. They start at zero. Mm-hmm. Even if beforehand it was already something detected. Yeah. Because the- Mm the rule makers, they decided it has to be this way, yeah.

Wojciech Wegrzynski 48:40

Mm-hmm.

Mohcine Chraibi 48:40

They don't care. But that was a simple scenario, but still not so straightforward or trivial as you might expect.

Wojciech Wegrzynski 48:47

I- i- in your code, can I drop, uh, occupants like that to parts of the model, appear them? Probably there are tricks to do it.

Mohcine Chraibi 48:54

To make them appear?

Wojciech Wegrzynski 48:55

Yeah, yeah.

Mohcine Chraibi 48:56

Yeah, yeah, you can. You can generate. In JuPedSim you can generate agents- But- at some time during the simulation.

Wojciech Wegrzynski 49:01

Mm-hmm. So and I assume because it's all Python coded, I can-

Mohcine Chraibi 49:04

Exactly

Wojciech Wegrzynski 49:04

I can, uh, define my

Mohcine Chraibi 49:06

own- Something that you can do by yourself. Mm. Uh, I mean, even the, the web-based It... you can define a polygon.

Wojciech Wegrzynski 49:15

Mm-hmm.

Mohcine Chraibi 49:15

Starting area where you can distribute, uh, people at once, or you can have a flow. Mm. Like, I can- would like to distribute, uh, people with a certain flow from time zero to time 100 or something like that. Mm. This is also possible.

Documentation, JuPedSim 2.0, Closing

Wojciech Wegrzynski 49:30

The f- the final thing I, uh, that's on my list is documentation. You praised FDS EVAC for having great, uh, documentation. How are you making sure that the person in the end has all the knowledge that they need to, to, to use this tool? I

Mohcine Chraibi 49:43

think, uh, I mean, the documentation, as I, I said this in my talk, is, is for me is more important than the code itself.

Wojciech Wegrzynski 49:49

Okay.

Mohcine Chraibi 49:50

And, um, so from day one, everything we write, we update the documentation. Mm-hmm. So the documentation right now is still fluid because the code is changing, but we want to make sure that people j- just can come and then get started with scenario zero. Uh, okay, and then move on. And then if they have question, how does this model work or how does this routing mo- or the cognitive map or whatever, everything I talked about in the talk is already documented there. Mm-hmm. Like, there is more details with the parameters, the values and such. So if someone want to, to, to dive in, it's There is the, the, the le- level one is just I want to use it, and then there are higher level where I want now to understand how the model is working

Wojciech Wegrzynski 50:32

actually. Mm-hmm. A- and what about, does it reach all the way back to, to fundamental models the, that, that are used for, I don't know, movement or, or way finding or

Mohcine Chraibi 50:43

Those then are JuPedSim.

Wojciech Wegrzynski 50:45

Okay.

Mohcine Chraibi 50:46

So of course there is a link. So if I want to know how the social force mo- model works, I would go to JuPedSim. There is It's already- Mm documented too. And we have another website that is very interesting. It's called pedestrian dynamics.org-

Wojciech Wegrzynski 50:59

Mm-hmm

Mohcine Chraibi 51:00

where we document all this basic knowledge, like what models they are, what routing. And so actually to keep it general for everybody- Mm and not like just for JuPedSim or just for PyFDSEvac.

Wojciech Wegrzynski 51:14

Fantastic. Were you going to Helsingborg next week?

Mohcine Chraibi 51:17

Have what?

Wojciech Wegrzynski 51:17

Are you going for the human, uh, behavior in, uh,

Mohcine Chraibi 51:20

fire? In two weeks I will go there. Yeah, yeah. Yes. Really, I, I'm looking forward because I would like to learn lots, uh, more about fire. Yeah.

Wojciech Wegrzynski 51:26

I, I It looks like a very, very promising event. I'm, I'm, I'm a- unfortunately not, uh, not attending, but, uh-

Mohcine Chraibi 51:32

You are missing out.

Wojciech Wegrzynski 51:33

I'm missing out on a lot- Yes. Yes but, uh, th- this is for me an, uh, a way to cope with, with this, uh- Yes by having-

Mohcine Chraibi 51:40

Sure. Really, see, I think it's very, gonna be a very nice, uh, conference. Mm. And human behavior and fire is something really I would like to learn more about. This is why I, I

Wojciech Wegrzynski 51:49

go there. Fantastic. So, uh, final thoughts. W- what's coming in the release to zero and, and, uh, what's then the next thing you start working on with, with- So

Mohcine Chraibi 51:58

JuPedSim 2.0 is gonna be three-dimensional.

Wojciech Wegrzynski 52:01

Three-dimensional.

Mohcine Chraibi 52:02

So that means we will be able to simulate, um, buildings.

Wojciech Wegrzynski 52:06

Mm-hmm.

Mohcine Chraibi 52:06

We will be able to simulate buildings with a new routing engine, so based on flow fields, also on graphs, so you can have both worlds. And more importantly, from my perspective as a modeler, we will have multi models. So that means you Agents can switch their- Okay model during the simulation-

Wojciech Wegrzynski 52:24

Mm-hmm

Mohcine Chraibi 52:25

which is something to-

Wojciech Wegrzynski 52:26

To a most appropriate model for the location they are.

Mohcine Chraibi 52:28

Exactly.

Wojciech Wegrzynski 52:28

Okay.

Mohcine Chraibi 52:29

That's what will come, and I think I'm very excited about this feature in specific.

Wojciech Wegrzynski 52:33

And it will work with, uh, PyFDSEvac?

Mohcine Chraibi 52:36

Of course. Always.

Wojciech Wegrzynski 52:37

Oh,

Mohcine Chraibi 52:37

always.

Wojciech Wegrzynski 52:37

Always. Always. Okay.

Mohcine Chraibi 52:38

PyFDSEvac- uses JuPedSim no matter what state it is in.

Wojciech Wegrzynski 52:43

Fantastic. Well, l- looking forward to that. Links are in the show notes. Yeah. Uh, your webpage is in the show notes. If anyone wants to, to learn more, uh, please visit those. And, uh, and, uh, yeah, keep modeling.

Mohcine Chraibi 52:54

Thanks for hosting me, Wojciech.

Wojciech Wegrzynski 52:57

And that's it. Links are in the show notes. Uh, resources are in the show notes. Uh, the whole package of FDS PyEvac or PyFDS Evac is available online. Uh, you can download it the, the way you install Python libraries, but you can also get some help from the websites that, that are linked. So you're very welcome to, to try, and I'm really glad we, we have a piece of software like that. Not, not that anything fundamentally wrong with the evacuation softwares that are available for engineers every day. But I think an open package which allows you to fiddle with the engine and, you know, change stuff in it, uh, is, is quite useful for the purpose of research and likely useful for different applications in, um, performance-based engineering where multiple simulations have to be carried and, and you would like to test some specific hypothesis in a, in a simple way. the integration of this with the FDS environment through, uh, the FDS wrapper libraries and the visibility library makes it straightforward from the perspective of, of implementing fire. I, like I'm not as optimist, uh, as, uh, Mohcine about, uh, the, easiness of this process. It's still like, uh, feels like programming, but at the same time, you know, with the, era of AI and chatbot support, you can do anything you want. So I think it's, it's very well manageable to, to get this running in, in your project and actually do very interesting things with it. Mohcine shown in his presentation some interesting things about, uh, what you can do with, with the model. it's an evacuation model embedded within your fire simulation that, that opens a lot of doors. I know, uh, softwares can do that, but, this is an open package, and we should all benefit from it. Um, very thankful for Ulisch to, to, to do those developments. Very thankful to them for, for supporting them. I think it's great a large institution is involved in, uh, developing and maintaining such a key items of, of software because this reassures me that it's, it's gonna be long-lasting effort. You know, it's not gonna disappear in a year before-- because, uh, project money runs out. It's, it seems to be a longer, uh, vision for this. So I hope it, it works for a long time, and I hope it's, it's improved, developed, and and put into use by a lot of colleagues. I think that will be it for today's episode. Uh, as Mohcine said, there's a Helsingborg Human, uh, Behavior in Fire workshop in, in just two weeks. So perhaps in this month, uh, we'll have more human evacuation topics. I, I, I like that theme perhaps right now. So, uh, we'll see about that. But, uh, one thing you can be sure, there's definitely more fire science coming your way, and that's gonna happen next Wednesday. So, uh, thanks for being here with me. Cheers. Bye.