Decision Support System for people evacuation: mobility demand and transportation planning,...

1Progetto Progetto SICUROSICURO

Decision Support System Decision Support System for people evacuation: for people evacuation: mobility demand and mobility demand and

transportation planningtransportation planningF. Russo, C. Rindone, G. Chilà

Università Mediterranea di Reggio Calabria

XV Convegno Nazionale SIDT Rende, 9-10 giugno 2008INPUT 2010, Potenza, 13-15 Settembre 2010INPUT 2010, Potenza, 13-15 Settembre 2010


I INTRODUCTION

II SW and DSSII.1 DEMANDII.2 EMERGENCY PLANNING

III CONCLUSIONS

CONTENTSCONTENTS


I INTRODUCTION


III CONCLUSIONS

CONTENTSCONTENTS


study

in-de

pth

Time

Space

STRATEGIC TACTIC OPERATIVE

NAT

ION

ALREG

ION

ALLO

CAL

DIREC

TIONAL

PRAT

ICAB

LEFE

ASIB

ILE

Planning dimensions

I INTRODUCTIONI INTRODUCTION


Invariante

Locale

Regionale

Nazionale


NATIONAL

REGIONAL

LOCAL

DIREC

TIONAL

PRAT

ICAB

LE

FEAS

IBLE

Time

Space

study

in-de

pth

External planning process



Invariante

Locale

Regionale

Nazionale


NATIONAL

REGIONAL

LOCAL

DIREC

TIONAL

PRAT

ICAB

LE

FEAS

IBLE

Time

Space

study

in-de

pth

Internal planning process



Objectives Constraints

Present situation

Strategies

Verify Objectives Constraints(EX ANTE)

Plan – Product

Future situation

Alternative scenarios SYSTEM O

F MO

DELS

Verify Objectives Constraints(EX POST)

Indicators (EX ANTE)

Indicators (EX POST)

Internal planning process



R = p V NP probability

N exposure

V vulnerability

Risk components (Russo, Vitetta, 2007)



Calamitous event

Preventiveinterventions

Calamitouseffects

Time

On goinginterventions

{Supply design {Demand management

Time for interventions


Decision Support System (DSS) and Software (SW) could assist decision makers before and during a calamitous event


Source: Homeland Security Exercise and Evaluation Program (HSEEP), 2009

USA approach to emergency planning modelling

PLANNING DEVELOPMENT

TRAININGIMPROVEMENT

ACTIONS

EXERCISES

Ex ante evaluations

Ex post evaluations



Source:

2009


USA approach to emergency planning modelling



Logical Framework Approach (LFA) and Project Cycle Managment

Source: European Commission, 2004

Ex ante evaluations

Ex post evaluations


P

PT

1) National law obliges Local Public Authority (P) to adopt emergency local plan

1) Local Public Authority (P) assign to Technician (T) activities to draw

2) The Technician (T) presents a proposal of plan to Local Public Authority (P)

P

4) Local Public Authority (P) submits to population (A) the proposal of plan

5) On the basis of remarks, Local Public Authority (P) approves plan

A

Actual emergency planning process

P: Political OrgansT: Technical OrgansA: Others Organs



Objectives Constraints

Present situation

Strategies

Verify

Plan – Product

Future situation

Alternative Scenarios

SYSTEM O

F MO

DELS

Indicators

P

PT1

P A

PT2

results of SICURO project

System of Models in evacuation planning process

P: Political OrgansT1: Technical Organs (planner)T1: Technical Organs (analyst)A: Others Organs



LOGFrame

LFA_INPUTSIF AND

LFA_ACTIVITIES IF AND

LFA_OUTCOMES IF AND

LFA_GOALS

THEN

LFA_OUTPUTS IF ANDTHEN

THEN

THEN

IndicatorsMeans of

verificationPlan descriptionExternal factors

Logical Framework Approach (LFA) for internal evacuation planning processLogical Framework Approach (LFA) for internal evacuation planning process



Results of Safety of Users in Road Evacuation (SICURO Project) System of Models in evacuation planning process

Demand

Single Building

Supply

SupplyDemand

Emergency vehicles

Refuge’s area

Trips for categoriesof users, modes

and refuge’s areas

Evacuation times of singles buildings

Evacuation times of population

Evacuation times of weak users

Access times on refuge’s area

Present situation

(PR

EV

EN

TIV

E –

ON

GO

ING

) IN

TE

RV

EN

TIO

NS

Demand

Single Building

Supply

SupplyDemand

Emergency vehicles

Refuge’s area

Trips for categoriesof users, modes

and refuge’s areas

Evacuation times of singles buildings

Evacuation times of population

Evacuation times of weak users

Access times on refuge’s area

Future situation



toward refuge’s area free (with or without user information on the system configuration) or targeted;

with different choice sets in relation to the alternatives: pedestrian, car, emergency vehicles, bus.

In ordinary conditions, the transportation demand can be simulated using the following sub - model:

Generation

Departure time

Distribution

Modal split

Route choice

with immediate or delayed approach, in relation to the time-gap available between t1 and t3

with free (with or without user information on the system configuration) or targeted departure;

free (with or without user information on the system configuration) or targeted.

In emergency conditions, the transportation demand can be simulated using the following sub - model:



delayedimmediate

t0 initial instant at which we decide to plan;t1 time at which the time when the dangerous event will happen is known or supposed

forecasted; t2 time at which the threat occurs in the system; t3 time at which no evacuation action is possible;t4 time at which the dangerous event ceases its effects on the system.

Russo, Vitetta (2007)

Effect on the population

Mitigation actions

Possible Not possible

YesNot

time∆1

EFFECT IN THE TIME

e.g. time bombe.g. time bomb tsunamitsunamie.g. earthquakee.g. earthquake

∆3

Different demand models have to be specified, in relation to event types, which can be classified according to their effects in space and in time.

e.g. earthquakee.g. earthquake



t0 time at which an hypothetical public decision maker decide to plan an evacuation from a considered area;

t1 time at which it’s possible to know when the hurricane will be in the considered area;

t2 time at which the hurricane reach the considered area ;

t3 time at which the hurricane starts its effects;

t4 time at which the hurricane ceases its effects on the population.

EXAMPLE: THE HURRICANE CASE

time∆1 ∆3

∆0 ≠0; ∆1 ≠0; ∆2 ≠0; ∆3 ≠0; ∆4 ≠0



EXAMPLE

Example ∆0 ∆1 ∆2 ∆3

Tsunami ≠0 ≠0 ≠0 ~0

Hurricane ≠0 ≠0 ≠0 ≠0

Twin Towers ≠0 ≠0 ≠0 ≠0

Bomb ≠0 ≠0 ~0/≠0 ~0

Cistern ≠0 ≠0 ≠0 ≠0

Chemistry pollution ≠0 ≠0 ≠0 ≠0

Vulcan eruption ≠0 ~0 ≠0 ≠0

Earthquake ≠0 0 0 0



A multy step approach to simulate demand in evacuation condition: user decisions & submodels

To evacuate or not?

By which transport mode?

Towards which destination?

By which path?

When?

GENERATION

DEPARTURE TIME

DISTRIBUTION

MODAL SPLIT

PATH CHOICE




To evacuate or not?



By which path?

When?

GENERATION

DEPARTURE TIME

DISTRIBUTION

MODAL SPLIT

PATH CHOICE

evacuation participation rates of evacuation zones

response curve, sensitive to the characteristics of the hurricane, time of day, type and timing of evacuation order

series of binary choices over time estimating a joint decision, generation with departure time, in the face of an oncoming hurricane

statistical approach (means and distributions)




To evacuate or not?

When?

GENERATION

DEPARTURE TIME

Sequential binary logit model

t,neU

t,nneU

)0UU0UUPr()UUUUPr(P 't,ee

't,nne

t,nne

t,ne

't,ee

't,nne

t,nne

t,ne

t >−≥−=>≥='t,n

e't,n

ne't UUU −=∆

])0UUPr(1[)0UUPr(P1t

1't

't,nne

't,ne

t,nne

t,ne

t ∏ ≥−−≥−=−

=)0UUPr( 't,n

ne't,n

e ≥−

Let and be the utility of a household n choosing to evacuate and not to evacuate, respectively, in time interval t, provided that the t interval was reached without evacuation. According to the random utility theory, the probability of a household evacuating in time interval t, ∀ t’≠t, is:

If the utility difference terms are independent in t, then:

where ], t’=1,2,…t, is the conditional probabilities of a household to evacuate in time interval t’ respectively, provided that the household has not evacuated yet




To evacuate or not?



By which path?

When?

GENERATION

DEPARTURE TIME

DISTRIBUTION

MODAL SPLIT

PATH CHOICE

disaggregate choice model for hurricane evacuation developed with post hurricane Floyd survey data collected in South Carolina in 1999




To evacuate or not?



By which path?

When?

GENERATION

DEPARTURE TIME

DISTRIBUTION

MODAL SPLIT

PATH CHOICE

disaggregate choice model for hurricane evacuation developed with post hurricane Floyd survey data collected in South Carolina in 1999

multinomial logit model to investigate the effect of risk areas in the path of a hurricane, and destination socioeconomic and demographic characteristics on destination choice behaviour.

path choice for emergency vehicle




To evacuate or not?



By which path?

When?

GENERATION

DEPARTURE TIME

DISTRIBUTION

MODAL SPLIT

PATH CHOICE

generation sub-model gives the level of demand in the study area according to the reference period and the population category

modal split sub-model gives the number of people using a given transport mode from a certain origin to a certain refuge area

distribution submodel gives the number of people choosing a given refuge area

SICURO RESEARCH PROJECT



Distribution

Generation Modal choice

E ,r E k,r k kkd (h,c,m) d (h) p (c) p (m / c)

ξ ξ= ⋅ ⋅∑

Evacuation demand modelEvacuation demand model

Modal choice with distribution

Modal choice

Distribution E,r Ek,r k kkd (h,c,m) d (h) p (m) p (c /m)= ⋅ ⋅∑

E,r Ek,r kkd (h,c,m) d (h) p (mc)= ⋅∑

Residents

Occasional customers

Employees

Weak user

Teaching and student

EFFECT IN THE SPACE

EFFECT IN THE TIME



RP data are not available for all dangerous events

models specified for hurricane evacuation, which are derived from observation of past evacuation behaviour, cannot be directly applied to other dangerous

events

prediction of user behaviour becomes essential, by:

evacuation trials SP (stated preference) surveys

RP data affected by the laboratory effect, like SP surveys with physical

verification

RP and SP approachesRP and SP approaches



SP surveys allow us to simulate several emergency scenarios

SP surveys must be designed, defining:

Proposed scenarios must be realistic and clear, in order to limit distortions between real and stated behaviour.

In light of such considerations, SP surveys play a very important role and RP surveys during evacuation

trials may be viewed as physical checking SP data.

EmergencyEmergency scenariosscenarios

Attributes for each Attributes for each scenarioscenario

Variation in level of Variation in level of attributesattributes

Choice mechanismChoice mechanism

Period of referencePeriod of reference

TargetsTargets

Effects in time and in spaceEffects in time and in space

for each user category

RP and SP approachesRP and SP approaches



In some evacuation conditions, the use of dynamic models is suggested.

Any Software (SW) or Decision Support System (DSS), for us knowledge, deals with this problem. We refer to Russo and Chilà (2007/c, 2008/b, 2010/a, 2010/b) for an analysis more complete of sequential dynamic approach (Gottman and Roy, 1990; Bakeman and Gottman, 1997).




In this paper DSS and SW are analysed : •to specify, to calibrate and to apply demand model in evacuation conditions

•to support transportation planning in evacuation conditions


I INTRODUCTION

II SW and DSSII.1 DEMANDII.2 PLANNING

III CONCLUSIONS

CONTENTSCONTENTS


SW AND DSS FOR DEMANDSW AND DSS FOR DEMAND

ALOGITALOGIT

HIELOWHIELOW

MMLM swMMLM sw

Some software used for demand model calibration


ALOGITALOGIT


PREPAREPREPARE ESTIMATEESTIMATE APPLYAPPLY

the logit model is set up and the data are

prepared and checked

unknown coefficients appearing in the

model are estimated from the data

the model is tested and/or applied for

forecasting

The last version of Alogit allows the parameter calibration with revealed preference or stated preference data.



HIELOWHIELOW


It allows a multinomial or a hierarchical (nested) logic model to be estimated.

To improve the quality of estimated models, HieLoW provides the analyst with detailed statistical information.

Based on recently developed trust-region methods, the maximization algorithm of HieLoW explicitly exploits, when

needed, the non-concavity of the loglikelihood function. A tutorial helps beginners get familiar with HieLoW.

A glossary and a permanent contextual help system are also included to facilitate the user's work.



MMLM swMMLM sw


The software realized by Train includes the following files:

mxlmsl.m is the code that the user runs; the user specifies the model within this code; doit.m is a script that is called up at the end of mxlmsl.m; it checks the data,

transforms the data into a more useful form, performs the estimation and prints results;check.m is a function that checks the input data and specifications; it provides error

messages and terminates the run if anything is found to be incorrect;loglik.m is a function that calculates the log-likehood function and its gradient; this

function is input to Matlab's fminunc command, which is part of Matlab's Optimization

Toolbox; this function calls llgrad2.m;llgrad2.m is a function that calculates for each person the probability of the chosen

alternatives and the gradient of the log of this probability;



MMLM swMMLM sw


The software realized by Train includes the following files:

der.m is a function that calculates the derivative of each random coefficient with respect

to the model parameters;makedraws.m is a function that creates the standardized draws that will be used in the

run, based on the specifications given by the user in mxlmsl.m; trans.m is a function that transforms the standardized draws into draws of coefficients;data.txt is an ascii file of data on vehicle choice; the data and its format are described

within mxlhb.m;myrunKT.out is the output file of running maxlhb.m with no modifications.

This software allows parameter calibration considering RP and SP surveys.

SW AND DSS FOR DEMAND SW AND DSS FOR DEMAND


Several commercial Decision Support Systems (DSS) are available to

evaluate transport demand.

Generally, these belong to the GIS (Geographic Information Systems)

software class.

GIS software integrates maps with their respective information or attributes.

Through its ability to link spatial data (maps) and non-spatial data (attribute

information) in one location, GIS provides a framework for efficient data

storage and data retrieval, intuitive display of information in a spatial

context, and combining various types of information so that the data may be

analyzed further.

Referring to demand model evaluation, GIS can be subdivided into two

main classes:

GENERIC GIS GENERIC GIS SOFTWARESOFTWARE

TRASPORTATION GIS TRASPORTATION GIS SOFTWARESOFTWARE



GENERIC GIS GENERIC GIS SOFTWARESOFTWARE

generic GIS software are developed and implemented in several fields (marketing,

planning, business analysis, transport, and so on)

Among the software belonging to the first class, we recall:

MapInfo, a powerful Microsoft Windows-based mapping and geographic analysis application from experts in location intelligence.

ArcInfo, the first GIS software available on the market






TRANSPORTATION TRANSPORTATION GIS SOFTWAREGIS SOFTWARE

OmniTRANS, which provides a versatile working

environment for multimodal transport planning and modelling; it offers an

integrated software platform for visual display of models

and graphical presentation of results, strong project

management tools to assist in managing all of the information

associated with model scenarios.






Emme/2, which is a graphical software tool for multimodal

transportation planning, which allows the transportation network to be modelled and assigns the traffic

generated under a given set of conditions







PTV, which is a software suite for transportation planning and operation

analyses






Among the software belonging to the first class, we recall: TransCAD, designed specifically for use by transportation professionals to store, display, manage, and analyze transportation data; TransCAD combines GIS and transportation modelling capabilities in a single integrated platform, providing capabilities that are unmatched by any other package; TransCAD can be used for all modes of transportation, at any scale or level of detail.TransCAD provides:a powerful GIS engine with special extensions for transportation; mapping, visualization and analysis tools designed for transportation applications; application modules for routing, travel demand forecasting, public transit, site location and area management;




DSS belonging to the second class generally include comprehensive tools for:

TRIP GENERATIONTRIP GENERATION

TRIP DISTRIBUTIONTRIP DISTRIBUTION

MODE SPLITMODE SPLIT





The goal of trip generation is to estimate the number of trips, by purpose, that are produced or

originate in each zone of a study area. Trip generation is performed by relating frequency of trips to

the characteristics of the individuals, the zone and the transportation network.


TRIP GENERATIONTRIP GENERATION

In some cases, there are two primary tools for modelling trip generation:

Cross-Classification, which separates the population in an urban area into relatively homogeneous groups based on certain socio-economic characteristics; average trip production rates per household or individual are then empirically estimated for each classification;

Regression Models, which allow evaluation and application of multivariable aggregate zonal models and disaggregate models at the household or individual level.




Trip distribution models are used to predict the spatial pattern of trips or other flows

between origins and destinations. DSS provide numerous tools with which to perform trip

distribution, including procedures to implement growth factor methods, apply previously-

calibrated gravity models, generate friction factors and calibrate new model parameters.


TRIP DISTRIBUTIONTRIP DISTRIBUTION



Mode choice models are used to analyze and predict the choices that individuals or groups of individuals make in selecting the transportation modes that are used for particular types of trips.

Typically, the goal is to predict the share or absolute number of trips made by mode. Software provides procedures for calibrating and applying mode choice models based on multinomial and nested logit models, and may be pursued at either a disaggregate or aggregate zonal level.

Estimation of the parameters in the nested logit and multinomial logit model is performed by the method of maximum likelihood, which calculates the set of parameters that are most likely to have resulted in the choices observed in the data.


Mode choice models are used to analyze and predict the choices that individuals or groups of individuals make in selecting the transportation modes that are used for particular types of trips.


MODE SPLITMODE SPLIT

SW AND DSS FOR DEMAND SW AND DSS FOR DEMAND


I INTRODUCTION


III CONCLUSIONS

CONTENTSCONTENTS


SW and DSS FOR EMERGENCY PLANNINGSW and DSS FOR EMERGENCY PLANNING

for project management

for specific component

generic to analyze transportation system in ordinary condition adopted for emergency condition

Logical Framework Approach Project Cycle Management

Project Cycle Management and Logical Framework Approach

specific to analyze transportation system in emergency condition



• Microsoft Project® (Microsoft, 2007)

• SmartDraw ® (SmartDraw.com, 2009) and Microsoft Visio® (Microsoft, 2007)

• Logical Decisions® for Windows, applied in Decision Analysis



Project Cycle Management

• Project Facilitator®, vers. 1.4 (Live Application Technology, 2004)




Logical Framework Approach

• LogFrame for Windows, vers. 1.0 (Maizemoor International, 2008)




Project Cycle Management and Logical Framework Approach

• TeamUP-PCM® (TEAM Technologies, Inc., 2008)

Stakeholder Analysis

Trees Analysis (Problem Analysis, Objectives Analysis)

Program and Project Structure

Conflict Analysis

Logical Framework

Schedule (WBS, Responsibility Matrix, Gantt Chart, CPM)

Performance Tracker (Output & Purpose Milestones)

Performance Budget




for specific component generic to analyze transportation system in ordinary condition adopted for emergency condition

– Macroscopic simulatione.g. EMME/2, TransCAD, VISUM, CUBE

– Mesoscopic simulatione.g. DYNASMART-P

– Microscopic simulatione.g. INTEGRATION, CORSIM



for specific component specific to analyze transportation system in emergency condition

• MASSVAC (MASS eVACuation)

• OREMS (Oak Ridge Evacuation Modeling System)

• ETIS (Evacuation Traffic Information System)

• HURREVAC (HURRicane EVACuation)

• SLOSH Model (Sea, Lake, and Overland Surges from Hurricanes)

• HAZUS-MH (Multi-Hazards U.S. Software)

• CATS (Consequence Assessment Tool Set) / Joint Assessment of

Catastrophic Events (JACE)

• MitigationPlan.com System

• Abbreviated Transportation Models (ATM)



I INTRODUCTION

II SW and DSSII.1 DEMANDII.2 PLANNING

III CONCLUSIONS

CONTENTSCONTENTS


CONCLUSIONSCONCLUSIONS

• Modelling tools implemented in SW and DSS to assist decision makers in preparing evacuation plans

• Relevant role of SW and DSS for transportation planning in emergency conditions

• real time• management of evacuation procedures


CONCLUSIONSCONCLUSIONS

• Advancements on mobility demand analysis and transportation planning

• Availability of SW and DSS for mobility demand and transportation planning

• SW and DSS for emergency planning • to collect and to represent data• to analyse relationships among data (models)

possibility to internalise the new negotiation among users, planners deciders and collectivity

• Future objectives• to develop structured planning and evaluation processes• to develop procedure and DSS able to simulate evacuation conditions

considering a dynamic sequential approach


Decision Support System Decision Support System for people evacuation: for people evacuation: mobility demand and mobility demand and

transportation planningtransportation planningF. Russo, C. Rindone, G. Chilà

Università Mediterranea di Reggio Calabria

XV Convegno Nazionale SIDT Rende, 9-10 giugno 2008INPUT 2010, Potenza, 13-15 Settembre 2010INPUT 2010, Potenza, 13-15 Settembre 2010


III.1 Definition of Evacuation Scenario and of Area of Study

III.2 Survey

III.3 Real Experimentation

III.4 Calibration and Validation of Proposed Models

III EXPERIMENTATIONIII EXPERIMENTATION


Area [m2] 35,300,000

Population 10,483

Workers 2,432

Melito Porto Salvo: Municipality of province of Reggio Calabria (Italy)




Invariante

STRATEGIC TACTIC OPERATIVEDIR

ECTI

ONAL

PRAT

ICAB

LE

FEAS

IBLE

NATIONAL

REGIONAL

LOCAL

LCPP

TIME

Space

study

in-de

pth

Melito Porto Salvo: Local Civil Protection Plan (LCPP)




• Simulation scenarios:

• a work day: 8.00-12.00

• a tank transporting dangerous goods is leaking (event type a)

• Mayor decides that surrounding area must be evacuated

• Categories of users in the evacuation area

• residents

• employees

• occasional customers

• teachers and students

• weak users

Melito Porto Salvo: verification applying system of models




Area [m2] 42,990

Residents 255

Employees 225

Melito Porto Salvo: area to test evacuation plan




Study area Zoning

Town of Melito Porto Salvo Study area

Area (km2) 35.30 0.04Resident 10483 255Employee 2432 225

Zone Area (m2) Zone Area (m2)

1 5091.50 7 3191.04

2 2869.96 8 3559.65

3 4064.14 9 5119.63

4 4885.16 10 3629.22

5 3801.01 11 3478.71

6 3300.35

TOTAL (m2) 42990.37

Urban area of Melito Porto Salvo - Province of Reggio Calabria (Italy)

Residential building

Public building

Mixed building




SURVEY

PRE-TEST

TEST

In order to calibrate the model we have carried:

directed to know socio-economic properties of the studying area

where an area with only public offices and one school is evacuated

where all the area are evacuated

directed to estimate habitual present user number and willingness user to evacuate

Revealed preferences before real

experimentation (RP)

Stated preferences before real experimentation (SP)

The data are recorded for the laboratory analysis. During the experimentation information have been founded with manual/automatic tools, 30

video cameras and interviewing evacuated user. From these surveys we can obtain variables for calibrating models.

DEMOGRAPHIC SURVEY AND CLASSIFICATION

directed to estimate habitual present user number and willingness user to evacuate

Revealed preferences during real

experimentation (SP with phisical check)


III.2 Survey


SURVEY – RP DATABASE

SchoolsPublic activitiesPrivate activitiesFamilies

Buildings

Scholl staff and pupils

Public sector employees

Private sector employees

Residents

Sex; Age; Profession;Weak user or not;

Driving licence; Vehicle possession;

Habitually present in the morning hours;

Willing to evacuate or not

Number component

Number employees and occasional customers

Number employees and occasional customers

Number teachings and students

AdressNumber floorsNumber exit

Type














III.2 Survey


Activity Number Activity Number

Clothes shop 4 Private office 10

Electronis, electrotechnis, mechanics, chemistry, car 6 Finance, insurance, credit 4

Food 8 Sport, free time, culture 5Agency 1 Public office 27Furniture 3 School 1Medicine and beauty 3Totale 72

Activity analysis

Socio-economic analysis

Building type Building number

Regular population

Occasional population

User category

Residential 23 89 0 Resident and weak user

Public

School 1 159 0 Teachers, pupils and weak user

Town hall 1 82 60Employee, occasional customer and

weak userCourt 1 7 3other 3 21 8

Mixed 28 262 99 Resident, weak user, employee and occasional customer

SURVEY – RP DATABASE


III.2 Survey


Resident analysis

36%

43%

21%

Present Not present Not found/interviewed/contacted

Worker analysis

14%

41% 45%

Present Not present Not interviewed/contacted

User categories Percentage of residents (%)

Percentage of employees (%)

Present willing to evacuation simulation 8 34

Present not willing to evacuation simulation 13 11

Not present 36 14Not contacted 8 14Not interviewed 14 27Not found 21 0

SURVEY - SP DATABASE


III.2 Survey


• 12/01/2007

Pre-Test

Evacuation of the school and of public buildings

• 1/03/2007

Test

Evacuation of the area of study


III.3 Real Experimentation


MAXIMUM LIKELIHOOD

In Maximum Likelihood estimation the value of the unknown parameters are obtained by maximising the probability of observing the choices made by a sample

of users.

LEAST SQUARES

For given observed data, the least squares values of model unknowns are the values minimizing the sum of

squared deviations, comparing the data to model predictions.

A simple, important example is bivariate linear regression, where a straight line is fitted to n pairs of

measurements on two variables, an independent variable and a dependent variable.

III. EXPERIMENTATON

CALIBRATION PARAMETER

Modal split

Distribution

Generation

Generation

RP

SP

SP with phisical check

SP

SP with phisical check



PARAMETER GENERATION MODAL SPLIT DISTRIBUTIONMODAL SPLIT

WITH DISTRIBUTION

General General For employee group

For employee group For employee group

SOCIO –

ECONOMIC

LEVEL OF

SERVICE

CALIBRATED PARAMETERCALIBRATED PARAMETER

XX XX

XX XX XX

XX





Socio-economic parameter Generation Modal split Distribution Modal split with distribution

Resident Not resident

General For employee

group

For employee group

For employee group

% of actives over residents X

% of students over residents X

% of housewifes over residents X

% of retired people over residents X

% of residents younger than 14 and older than 5 years X

% resident younger than 19 and older than 15 years X



% resident overthan 65 years X

family number X

employee number X

teaching and scholl number X

Weak user number X

Dummy for employees of age below 45 years XDummy equal to 1 if the employee’s level is higher than 2, 0 otherwise X X X

Dummy equal to 1 if the employee’s level is higher than 3, 0 otherwise X X

Dummy equal to 1 if the user is a women, 0 otherwise X X



Level of service parameter Generation Modal split DistributionModal split

with distribution

General General For employee group

For employee group

For employee group

Time on the pedestrian network from origine to the refuge’s area

X X

Time on the road network from origine to the refuge’s area X X

Distance as the crow flies between origine and refuge’s area X X

Distance on the pedestrian network between origine and refuge’s area

X

Dummy origine for zone 10 X


MAIN OUTCOMESMAIN OUTCOMES


Generation model Distribution Modal choice

Present in the area

Willing to evacuation simulation

Refuge’s are fixed

Refuge’s are not fixed

Car PedestrianBus or

Emergency vehicles

37% 39% 84% 16% 80% 20%

77% 75% 84% 16% 80% 20%

80% 67% 84% 16% 80% 20%

92% 100% 100% 100%

100% 100% 100% 100%

Residents

Occasional customers

Employees

Weak users

School staff and pupils



Spec. Parameters Value T-Student ρ2

1

αFL % of actives over residents 0.13 2.82

0.99αNS % of students over residents 0.33 4.74

αNC % of housewives over residents 1.00 42.36

αNR % of retired people over residents 0.14 2.55

2

αE2% of residents younger than 14 and older than 5 years -0.09 -0.28

0.96

αE3% of residents younger than 19 and older than 15 years

-0.83 -1.16


-0.32 -0.46


0.80 8.36

αE6 % of residents over 65 years -0.15 -0.51

3 mE,R Attendance coefficient 0.59 23.51 0.98

Calibration of resident generation model: PRESENT RESIDENT


Generation


User category λ (%)Employee 77Occasional customer 80School staff and pupils 92Weak user 100

Calibration of resident generation model: PRESENT NOT - RESIDENT

SP DATA RP DATAUser category ξ (%) ξ (%)Resident 39 /Employee 75 50Occasional customer 67 /School staff and pupils 100 100Weak user 100 100

Calibration of resident generation model: USER WILLING TO EVACUATE ON THE PRESENT


Generation

Generation


Calibration of modal split model: TEST 01/03/2007

Parameters Alt Specific.1

Specific.2

Specific.3

βD Distance as the crow flies between origine and refuge’s area

2 -0.0072

(-0.3)

βCar Dummy for car alternative 1 -0.8579

(-0.1)

βTRP Time on pedestrian network from origine to the refuge’s area

1 -0.2881 -0.6924 -0.3049

(-1.0) (-0.4) (-0.9)

βTRC Time on the road network from origine to the refuge’s area

2 -1.0590 -1.186 -0.9961

(-0.90) (-0.90) (-0.80)

Initial Likelihood -30.4985 -30.4985 -30.4985

Final Likelihood -28.3578 -28.3219 -28.3512

ρ2 0.07 0.07 0.07

Alternatives: 1 Pedestrian, 2 Car


Modal split


Calibration of modal split model: Town Hall Model, TEST 01/03/2007

Parameters Alt. Specific.1 Specific.2

βWE1 Dummy for employees of age below 45 years 1 2.5660 3.0710

(1.70) (1.90)

βDRC Distance on pedestrian network between origine and refuge’s area

1 -0.0025 -0.0027

(-1.60) (-1.90)

βL2 Dummy if the employee level is higher than 2, 0 otherwise 2 0.2238

(0.20)

βL3 Dummy if the employee level is higher than 3, 0 otherwise 2 1.1830

(0.70)

βCW Dummy equal to 1 if the user use the car to go to work, 0 otherwise

2 0.3494 0.3042

(0.60) (0.50)

βWomen Dummy equal to 1 if the user is a women, 0 otherwise 1 1.7770 1.9390

(1.30) (1.30)

Initial Likelihood -14.56 -14.56

Final Likelihood -7.95 -7.68

ρ2 0.45 0.47Alternatives: 1 Pedestrian, 2 Car


Modal split


Calibration of distribution model: Town Hall Model, TEST 01/03/2007


Specific.2

βWomen Dummy equal to 1 if the user is a women, 0 otherwise

1 2.850

(3.3)

βL2 Dummy if the employee level is higher than 2, 0 otherwise

1 1.289 0.2657

(2.5) (0.4)

βz10 Dummy origine for zone 10 2 1.264 1.483

(2.8) (3.9)

Initial Likelihood -39.5094 -39.5094

Final Likelihood -34.1374 -25.8085

ρ2 0.14 0.35

Alternatives: 1 Refuge’s area fixed by public decision maker (RAF), 2 Other refuge’s area (RAO)


Distribution


Calibration of modal split with distribution model: Town Hall Model, TEST 01/03/2007


βTRP,RA1 Time on pedestrian network from origine to the refuge’s area 1

1 -0.2688

(-1.4)

βTRP,RA2 Time on pedestrian network from origine until to refuge’s area 2

2 -1.0260

(-1.5)

βTRC,RA1 Time on road network from origine to the refuge’s area 1

3 -1.9670

(-1.3)

Initial Likelihood -40.6487

Final Likelihood -33.5616

ρ2 0.17

Alternatives: 1 Pedestrian with refuge’s area 1; 2 Pedestrian with refuge’s area 2; 3 Car with refuge’s area 1


Modal split

Distribution

Decision Support System for people evacuation: mobility demand and transportation planning,...

Documents

Transcript of Decision Support System for people evacuation: mobility demand and transportation planning,...