utsp seminar

Upload: jayashree-arumugam

Post on 06-Apr-2018

215 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/3/2019 Utsp Seminar

    1/17

    STOCHASTIC USEREQUILIBRIUM

    SUBMITTED BY

    SUDHA DAS KHAN 11ID60R18

    JAYASHREE.A 11ID60R24

  • 8/3/2019 Utsp Seminar

    2/17

    TRAFFICASSIGNMENT

    Traffic assignment, the selection of routes

    (alternative called paths) between origins and

    destinations in transportation networks.

  • 8/3/2019 Utsp Seminar

    3/17

    TRAFFIC ASSIGNMENT MODEL

    Traffic assignment model has three submodels:

    1. Supply model: simulates variation in network

    performance due to user behaviour

    2. Demand model: simulates variation in user

    behaviour due to network performance

    3. Supply-demand interaction model: simulates the

    user network interaction

  • 8/3/2019 Utsp Seminar

    4/17

    TRAFFIC ASSIGNMENT BASED ON

    1. Dynamic approach(within day or day to day)

    2. Utility approach(deterministic or stochastic)

    3. Service regularity(regular or irregular)

    4. User information system(present or not)

    5. User type(frequent or occasional)

  • 8/3/2019 Utsp Seminar

    5/17

    UTILITY THEORY

    The utility to a individual traveler offered by given travel choice alternativedepends on:

    speed

    safety cost

    travel time

    translated into its monetary value of traveler.

    The traveler wishes to maximize his/her utility

    DETERMINISTIC UTILITY THEORY

    o The utility of a traveler is the function of travel choice alternatives.

    o It is assumed that the function remains constant for a particular route or

    for a particular class of people.o Uij = bij cij tij

    o

    It is not possible to predict with certainty the alternative that the genericdecision maker will select

    Uij utility received from trip made between i and jbij income of the travelercij cost of travel - parameter which converts time into monetary value

  • 8/3/2019 Utsp Seminar

    6/17

    STOCHASTIC UTILITY THEORY

    The utility function takes into account the individual strength of preference for

    route choice

    It is not possible to predict the alternative of the traveler with certainty

    Hence it is expressed in terms of probability

    The perceived utility Uij, expressed as sum of systematic utility(Vij) and random

    residual(ij)

    Uij = Vij +

    ij

    Vji = E[U

    ij]

    2i,j = Var[Uij ]

    And therefore,

    E[Uij] = Vj

    i Var[Vj

    i

    ] = 0

    E[ij] = 0 Var[ij] =

    2i,j

    The probability is given by,

    Pi[j/Ii] = Pr[Vij +

    ij > V

    ik +

    ik] = Pr[V

    ij - V

    ik >

    ik -

    ij]

  • 8/3/2019 Utsp Seminar

    7/17

    TRAFFIC ASSIGNMENT BASED ON UTILITYAPPROACH

    DETERMINISTICAPPROACH

    STOCHASTICAPPROACH

    It does not take into account theoverlapping of the path

    Simple structure and ease of use

    It takes into account theoverlapping at the cost of pathenumeration

    The drawbacks of modified logitmodel and taken careRequires Monte Carlo Simulation orof complete path enumeration andnumerical integration of the

    multivariate Normal distribution

    All or nothingassignment

    Method of

    Successive Average

    Modified Algorithm

    Capacity RestraintAssignment

    IncrementalAssignment

    Logit Model

    Modified LogitModel

    Probit Model

    STOCHASTICAPPROACH

    Logit Model

    Modified LogitModel

    Probit Model

  • 8/3/2019 Utsp Seminar

    8/17

    STOCHASTIC USER EQUILIBRIUM(LOGIT MODEL)

    WHICH PATH IS CHEAPEST;

    SHORTEST; LESSCONGESTED..

    STOCHASTICLOADING METHOD CAPACITYRESTRAINT

  • 8/3/2019 Utsp Seminar

    9/17

    The proportion of traffic between O-D pair(r,s) choosingto use route kwhich has a mean cost of Crs

    k is given by,

    the cost term are independent Gumbel variate

    is the spread parameter

    The model has Markovian nature, then

    B(j) set of before nodes for node j and cij is costof link

    exp( )

    exp( )

    k

    k rs

    jrs

    rs

    j

    CP

    C

    ( )

    exp( )rj ri ij

    i B j

    jW W C

    Stochastic Loading Method

  • 8/3/2019 Utsp Seminar

    10/17

    Flow on link(i,j) from origin ris given by

    Xrj outflow of traffic with origin rto nodej

    Satisfaction function, the mean perceived cost of

    travel between O-D pairs(r,s)

    exp( ) / rjij

    rj

    WijX X C

    1 1log( ) log exp( )

    k

    rs rs rsk

    S W C

  • 8/3/2019 Utsp Seminar

    11/17

    CAPACITY RESTRAINT

    Here capacity restraint in the form of link-cost flow

    function

    The link flow at iteration n+1 is

    xn+1=xn+n[y(n)-x(n)]

    n - step length (1/n+1) (Method of successive average)

    y(n) all or nothing assignment solution

  • 8/3/2019 Utsp Seminar

    12/17

    STOCHASTIC USER EQUILIBRIUMALGORITHM

    Step 1: Set up an initial flow pattern xij, typically by carrying out a

    stochastic loading based on free flow link costs

    Step 2:

    Compute new link costs c(n) based on the current flow

    pattern x(n)

    Carry out a stochastic loading using these link costs,

    producing an auxiliary flow pattern y(n)

    Using the output values of Srs and the vector of partial

    derivatives z/ xa, compute z0 and g0

    0

    ( ) ( ) ( ) ( )ax

    a a a a rs rs

    a a rs

    z x C x dx x c x q S x ( ) ( )( ) ( )

    ( ) ( )n na a

    a a a a

    a

    dz x dc xg x y y x

    d dx

  • 8/3/2019 Utsp Seminar

    13/17

    Step 3: Compute link cost based on auxiliary solution and

    use these to determine z1 and g1

    Step 4: using z0, z1 and g0, g1, estimate the optimal step

    length n, n = -g0/(-g0+g1)

    update the current solution,

    xn+1=xn+n[y(n)-x(n)]

    Step 5: at stochastic user equilibrium the auxiliary and

    current flow pattern are the same

  • 8/3/2019 Utsp Seminar

    14/17

    CONCLUSION

    The most widely used algorithm is Method of

    Successive Average

    Inspite of its drawbacks, the logit method is widelyused

    Many algorithm are also developed to take care of

    the drawbacks in logit model

  • 8/3/2019 Utsp Seminar

    15/17

    REFERENCE

    ALGORITHMS FOR LOGIT-BASED STOCHASTIC USEREQUILIBRIUM ASSIGNMENT MIKE MAHER - Department ofCivil and Transportation Engineering, Napier University,Edinburgh EH10 5DT, U.K.

    Transportation System Analysis: Models and Application byEennio Cascetta

    Transportation System Engineering: theory and method byEennio Cascetta

  • 8/3/2019 Utsp Seminar

    16/17

  • 8/3/2019 Utsp Seminar

    17/17

    EXAMPLE:

    Network consisting of three parallel links between asingle O-D pair. The link cost flow relations are:

    c1=3+x1 c2=2+2x2 c3=2.5+1.5x3

    and the demand is one.

    n z(x) x1(n) x2

    (n) x3(n)

    0 -1.6055308 0.1863237 0.5064804 0.3071959

    1 -1.6267193 0.2514581 0.4058328 0.3427091

    2 -1.6268206 0.2588580 0.4062035 0.3349384

    3 -1.6268212 0.2591917 0.4056907 0.3351176

    4 -1.6268212 0.2592326 0.4056927 0.3350746

    5 -1.6268212 0.2592345 0.4056899 0.3350756

    6 -1.6268212 0.2592347 0.4056899 0.3350754