Numerical simulation and optimization models for socio-dynamical features of crowd evacuation

Anton von Schantz

Numerical simulation and optimization models for socio-dynamical features of crowd evacuation

Julkaisun otsikon käännös: Väkijoukon evakuoinnin sosiodynaamisten ilmiöiden numeeriset simulointi- ja optimointimallit

Anton von Schantz

Matematiikan ja systeemianalyysin laitos

Tutkimustuotos: Doctoral Thesis › Collection of Articles

Abstrakti

Suurten joukkotapahtumien sekä julkisen liikenteen kasvu aiheuttaa turvallisuushaasteita erityisesti hätätilanteissa. Väkijoukon evakuoinnin ongelmien ratkaisemiseen on kehitetty laskennallisia malleja. Nykyään kyetäänkin numeerisesti simuloimaan erilaisia näihin tilanteisiin liittyviä fysikaalisia ilmiöitä realistisesti, mutta malleja, jotka kuvaavat yksilöiden käyttäytymisestä johtuvia ilmiöitä on vähemmän. Näiden lisäksi tarvitaan myös malleja, jotka johtavat käytännön ratkaisuihin katastrofaalisten tapahtumien ennaltaehkäisemiseksi. Tässä väitöskirjassa tutkitaan väkijoukon liikkeen matemaattista sekä laskennallista mallintamista ja yksilöiden käyttäytymisen vaikutusta haitallisten ilmiöiden syntymisessä. Tutkimme myös nopeimman opastetun poistumisen tehtävää. Eli kuinka keltaliivisten oppaiden tulisi ohjata väkijoukko rakennuksesta ulos minimiajassa? Nopeimman opastetun poistumisen tehtävää tutkitaan myös kun väkijoukon liikehdinnässä on epävarmuutta.Väitöskirjassa väkijoukko mallinnetaan monen agentin järjestelmänä. Sen dynamiikkaa kuvataan Newtonin dynamiikkaan perustuvalla sosiaalisen voiman mallilla ja agenttien käyttäytymistä evoluutiopeliteorialla. Mallia simuloidaan rakentamalla simulaatioympäristö, joka implementoidaan supertietokoneelle. Numeeriset simuloinnit osoittavat, että agenttien lokaalin käyttäytymisen seurauksena syntyy ei-monotoninen dynamiikka. Pullonkaulan läpi poistuessa väkijoukon takaosa käyttäytyy kärsimättömästi. Se työntää edessä olevia agentteja, mikä kasvattaa painetta etuosassa. Tämän seurauksena väkijoukkoon muodostuu agenteista koostuvia holvirakenteita, jotka tukkivat pullonkaulan sekä hidastavat evakuointia. Holvirakenteet murtuvat niihin kohdistuvien epävakaiden kuormien takia. Simulointitulokset yhtyvät kokeellisten evakuointitutkimusten kanssa. Nopeimman opastetun evakuoinnin tehtävään kehitetään uusia matemaattisia malleja sekä algoritmeja. Uudet menetelmät perustuvat matemaattiseen optimointiin, kuten skenaario-optimointiin, geneettisiin algoritmeihin, numeeriseen simulointiin pohjautuvaan optimointiin sekä monitavoiteoptimointiin. Lisäksi pahimmat mahdolliset skenaariot huomioidaan riskimitalla. Nopeimman opastetun evakuoinnin tehtävän ratkaisu antaa oppaiden lukumäärän, niiden lähtöpaikat sekä niille määrätyt uloskäynnit. Näytetään, että jos oppaiden määrä on rajoitettu, joudutaan tekemään kompromissi pahimman mahdollisen skenaarion optimoivan sekä keskimääräisesti optimaalisen evakuointisuunnitelman välillä. Kuitenkin jos on riittävästi oppaita, yksilöiden sekä väkijoukon liikkeen epävarmuuden vaikutus häviää. Väitöskirja tarjoaa useita uusia käytännön työkaluja väkijoukon evakuoinnin dynaamisten ominaisuuksien numeeriseen simulointiin ja optimointiin.

Julkaisun otsikon käännös	Väkijoukon evakuoinnin sosiodynaamisten ilmiöiden numeeriset simulointi- ja optimointimallit
Alkuperäiskieli	Englanti
Pätevyys	Tohtorintutkinto
Myöntävä instituutio	Aalto-yliopisto
Valvoja/neuvonantaja	Ehtamo, Harri, Vastuuprofessori Ehtamo, Harri, Ohjaaja
Kustantaja	Aalto University
Painoksen ISBN	978-952-64-0368-7
Sähköinen ISBN	978-952-64-0369-4
Tila	Julkaistu - 2021
OKM-julkaisutyyppi	G5 Tohtorinväitöskirja (artikkeli)

Tutkimusalat

väkijoukon evakuointi
monen agentin järjestelmä
evoluutiopeliteoria
numeerinen simulointi
matemaattinen optimointi
ei-monotoninen dynamiikka

Pääsy asiakirjaan

http://urn.fi/URN:ISBN:978-952-64-0369-4

SFX-saatavuus

Koko teksti

Science-IT
Mikko Hakala (Manager)
Perustieteiden korkeakoulu
Laitteistot/tilat: Facility

Siteeraa tätä

@phdthesis{53869f340ab044b5aed941914266f589,

title = "Numerical simulation and optimization models for socio-dynamical features of crowd evacuation",

abstract = "The rapid increase of various mass gatherings and overcrowded festivals pose serious challenges, for example in case of emergency. Computational models may help to address issues related to these socio-physical systems, and in particular evacuating crowds. Physics-inspired self-driven particle models can describe most of the physics of moving crowds. However, there is still a need for comprehensive crowd models that can describe collective crowd effects, starting from individual crowd members' decision-making. In addition to models being able to describe harmful crowd phenomena, they should also prescribe solutions to prevent them. This dissertation concerns the mathematical and computational modeling of an evacuating crowd. The main focus is on studying how individual decision-making causes the harmful physical effects in a bottleneck evacuation. How should rescue guides be used to minimize the evacuation time of a crowd? What is the effect of uncertain crowd movement patterns on the minimum time evacuation plan? A multiagent framework is used to model the crowd. Its members are modeled as agents that interact with each other. The crowd dynamics are described using social force model based on Newtonian dynamics, and the agents' decision-making is described using evolutionary game theory. The model is studied by developing a simulation environment, which is implemented in a high-performance computing cluster. Numerical simulations show that due to the locally-played game, non-monotonous dynamical effects emerge. In a bottleneck congestion, the back of the crowd behaves impatiently. It pushes the agents in front of it, and pressure increases. As a result, arch-like structures form, capable of interrupting the flow and slowing down the evacuation. The arches break down due to fluctuating loads. The results coincide with findings from behavioral and physical evacuation experiments. New mathematical models and algorithms are developed to solve the minimum time crowd evacuation problem with rescue guides. The new methods are based on mathematical optimization, namely, on scenario optimization, genetic algorithms, numerical simulation-based optimization, and bi-objective optimization. Also, worst-case scenarios are accounted for with a risk measure. The solution to the minimum time evacuation problem gives the number of guides, their initial positions, and exit assignments. It is shown that there is a tradeoff between the evacuation plan that performs well across scenarios, and the one that performs well on the worst-case scenario. With enough guides, the uncertainty in the individual and crowd movement patterns is mitigated. This dissertation provides new practical tools for numerical simulation and optimization of dynamical features of crowd evacuation, and hopefully gives ways to prevent fatal accidents in emergencies.",

keywords = "crowd evacuation, multiagent system, evolutionary game theory, numerical simulation, mathematical optimization, non-monotonous dynamics, v{\"a}kijoukon evakuointi, monen agentin j{\"a}rjestelm{\"a}, evoluutiopeliteoria, numeerinen simulointi, matemaattinen optimointi, ei-monotoninen dynamiikka, crowd evacuation, multiagent system, evolutionary game theory, numerical simulation, mathematical optimization, non-monotonous dynamics",

author = "{von Schantz}, Anton",

year = "2021",

language = "English",

isbn = "978-952-64-0368-7",

series = "Aalto University publication series DOCTORAL DISSERTATIONS",

publisher = "Aalto University",

number = "64",

school = "Aalto University",

}

TY - THES

T1 - Numerical simulation and optimization models for socio-dynamical features of crowd evacuation

AU - von Schantz, Anton

PY - 2021

Y1 - 2021

N2 - The rapid increase of various mass gatherings and overcrowded festivals pose serious challenges, for example in case of emergency. Computational models may help to address issues related to these socio-physical systems, and in particular evacuating crowds. Physics-inspired self-driven particle models can describe most of the physics of moving crowds. However, there is still a need for comprehensive crowd models that can describe collective crowd effects, starting from individual crowd members' decision-making. In addition to models being able to describe harmful crowd phenomena, they should also prescribe solutions to prevent them. This dissertation concerns the mathematical and computational modeling of an evacuating crowd. The main focus is on studying how individual decision-making causes the harmful physical effects in a bottleneck evacuation. How should rescue guides be used to minimize the evacuation time of a crowd? What is the effect of uncertain crowd movement patterns on the minimum time evacuation plan? A multiagent framework is used to model the crowd. Its members are modeled as agents that interact with each other. The crowd dynamics are described using social force model based on Newtonian dynamics, and the agents' decision-making is described using evolutionary game theory. The model is studied by developing a simulation environment, which is implemented in a high-performance computing cluster. Numerical simulations show that due to the locally-played game, non-monotonous dynamical effects emerge. In a bottleneck congestion, the back of the crowd behaves impatiently. It pushes the agents in front of it, and pressure increases. As a result, arch-like structures form, capable of interrupting the flow and slowing down the evacuation. The arches break down due to fluctuating loads. The results coincide with findings from behavioral and physical evacuation experiments. New mathematical models and algorithms are developed to solve the minimum time crowd evacuation problem with rescue guides. The new methods are based on mathematical optimization, namely, on scenario optimization, genetic algorithms, numerical simulation-based optimization, and bi-objective optimization. Also, worst-case scenarios are accounted for with a risk measure. The solution to the minimum time evacuation problem gives the number of guides, their initial positions, and exit assignments. It is shown that there is a tradeoff between the evacuation plan that performs well across scenarios, and the one that performs well on the worst-case scenario. With enough guides, the uncertainty in the individual and crowd movement patterns is mitigated. This dissertation provides new practical tools for numerical simulation and optimization of dynamical features of crowd evacuation, and hopefully gives ways to prevent fatal accidents in emergencies.

AB - The rapid increase of various mass gatherings and overcrowded festivals pose serious challenges, for example in case of emergency. Computational models may help to address issues related to these socio-physical systems, and in particular evacuating crowds. Physics-inspired self-driven particle models can describe most of the physics of moving crowds. However, there is still a need for comprehensive crowd models that can describe collective crowd effects, starting from individual crowd members' decision-making. In addition to models being able to describe harmful crowd phenomena, they should also prescribe solutions to prevent them. This dissertation concerns the mathematical and computational modeling of an evacuating crowd. The main focus is on studying how individual decision-making causes the harmful physical effects in a bottleneck evacuation. How should rescue guides be used to minimize the evacuation time of a crowd? What is the effect of uncertain crowd movement patterns on the minimum time evacuation plan? A multiagent framework is used to model the crowd. Its members are modeled as agents that interact with each other. The crowd dynamics are described using social force model based on Newtonian dynamics, and the agents' decision-making is described using evolutionary game theory. The model is studied by developing a simulation environment, which is implemented in a high-performance computing cluster. Numerical simulations show that due to the locally-played game, non-monotonous dynamical effects emerge. In a bottleneck congestion, the back of the crowd behaves impatiently. It pushes the agents in front of it, and pressure increases. As a result, arch-like structures form, capable of interrupting the flow and slowing down the evacuation. The arches break down due to fluctuating loads. The results coincide with findings from behavioral and physical evacuation experiments. New mathematical models and algorithms are developed to solve the minimum time crowd evacuation problem with rescue guides. The new methods are based on mathematical optimization, namely, on scenario optimization, genetic algorithms, numerical simulation-based optimization, and bi-objective optimization. Also, worst-case scenarios are accounted for with a risk measure. The solution to the minimum time evacuation problem gives the number of guides, their initial positions, and exit assignments. It is shown that there is a tradeoff between the evacuation plan that performs well across scenarios, and the one that performs well on the worst-case scenario. With enough guides, the uncertainty in the individual and crowd movement patterns is mitigated. This dissertation provides new practical tools for numerical simulation and optimization of dynamical features of crowd evacuation, and hopefully gives ways to prevent fatal accidents in emergencies.

KW - crowd evacuation

KW - multiagent system

KW - evolutionary game theory

KW - numerical simulation

KW - mathematical optimization

KW - non-monotonous dynamics

KW - väkijoukon evakuointi

KW - monen agentin järjestelmä

KW - evoluutiopeliteoria

KW - numeerinen simulointi

KW - matemaattinen optimointi

KW - ei-monotoninen dynamiikka

KW - crowd evacuation

KW - multiagent system

KW - evolutionary game theory

KW - numerical simulation

KW - mathematical optimization

KW - non-monotonous dynamics

M3 - Doctoral Thesis

SN - 978-952-64-0368-7

T3 - Aalto University publication series DOCTORAL DISSERTATIONS

PB - Aalto University

ER -

Numerical simulation and optimization models for socio-dynamical features of crowd evacuation

Abstrakti

Tutkimusalat

Pääsy asiakirjaan

SFX-saatavuus

Sormenjälki

Laitteet

Science-IT

Siteeraa tätä