lOMoARcPSD|2396987

Overwiev of spatial modelling

Transportation and Spatial Modelling (Technische Universiteit Delft)

Distributing prohibited | Downloaded by Usama Shahdah (usama.elrawy@gmail.com)

 lOMoARcPSD|2396987

4-step model

Distributing prohibited | Downloaded by Usama Shahdah (usama.elrawy@gmail.com)

 lOMoARcPSD|2396987

Distributing prohibited | Downloaded by Usama Shahdah (usama.elrawy@gmail.com)

 lOMoARcPSD|2396987

zonal data
household, counter,
data from

trip generation
atraction)
(production,
how many people go

trip distribution
people go
where exactly
od-matrix

modal split
choice (bus, car)
how people go, mode

period of day

assignment

Choice Probit Logit

Nested (decision) binary

Choice for each link Choice for route

Parameter β describes
sensitivity for differences:
zero: not sensitive
large: very sensitive (“all or
nothing”)

Distributing prohibited | Downloaded by Usama Shahdah (usama.elrawy@gmail.com)

 lOMoARcPSD|2396987

Matrix balancing algorithm

Distributing departures or arrivals over rows or columns based on:

• Older OD-matrix (i.e. growth factor)

• Willingness to travel (i.e. deterrence function)

• Choice probabilities

Two modelling methods

• Singly constrained: simply distributing departures over destinations or arrivals over origins

• Doubly constrained: iteratively distributing departures and arrivals

Trip generation (trip production/attraction)

Input – data from household, counter, zonal data

Output - How many people go (production, atraction)

Logit choice

Regression Cross-clasification Stop and repeat model

(binary logit)
(descrete choice model)

Distributing prohibited | Downloaded by Usama Shahdah (usama.elrawy@gmail.com)

 lOMoARcPSD|2396987

Data Hoesehol Household Household (cars,

Zone data income, family size)
- Does not give information Needs lots of data
about activity end
Do not the amount of
Not show trip purpose trips in 1 tour

Need lots of data

• No extrapolation beyond
the calibrated groups
• Large samples are required
(at least 50 obs. per group)
• What is the best grouping?
+ Can be done at • Groupings are
aggregate and independent of zone system
disaggregate level • Relationships do not need
to be linear
• Each group can have a
different form of relationship
Main characteristic Production calculated Production is calculated
using the probability of
trip(tour) rates
Describtive Describtive propabilistic
detarmenistic detarmenistic
Trip/tour Both Trips both

Can model atraction

and production for
non home end

Trip distribution
skim matrics (costs) , od-matrix (trips)

Two key principles in trip distribution

• Big produces/atracts more

• Nearby atracts more

Distributing prohibited | Downloaded by Usama Shahdah (usama.elrawy@gmail.com)

 lOMoARcPSD|2396987

input – production atraction

out put – od matrics, - where exactly people go

advantages disadvantages
Growth factor Multiply od-matrix on % -Network specific -New residential
peculiarities can be zones are difficult to
captured by capture
g-growth factor observations - Historical paterns
T-matrix cell (base and predicted) - A base matrix is may change over
more time
understandable and
verifiable
than a model
Gravity model Basic gravity model
(use if you Tij=pQiXiFij
have the
determinse Tij= # trips from zone i to zone j
function f©) p= measure of average trip intensity
Qi= production potential of zone j
Xi= atraction potential of zone i
Fij= willingness to travel from i to j

Doubly constrained gravity

model( both production and
atraction)

Tij – number of trips from I to j

ai-scaling factor for origin zone i
bj-scaling factor for destination zone j
Pi-number of trips produced by zone i
Aj-nember of trips atracted by xone j
Fij – acessibility of zone j from zone i
Cij-impedance from zone I to zone j
The acessability depends on the impedance
between zones I and j

Determine the values for the

multipliers Qi and Xj for the first
iteration of the 1. Since through
traffic is given, the values for the
production and attraction for the
zones north and South should be
reduced to in- en outgoing traffic
only.
Furthermore the travel costs for
through traffic should be set to 99999
(or the willingness to travel to 0), just
as for the intrazonal travel costs of
these two zones.
2. The rest of the matrix is filled with

Distributing prohibited | Downloaded by Usama Shahdah (usama.elrawy@gmail.com)

 lOMoARcPSD|2396987

the corresponding values of the

deterrence function.
3. The rows are summed and the
multiplier Qi is calculated as
production/sum row.
4. The rows are multiplied by the
corresponding Qi values.
5. The columns are summed and the
multiplier Xj is calculated as
attraction/sum column.

singly constrained (just one) – when

production is not equal to atraction
Entropy
maximization
Choice
modeling
For tours

TRij-number of tours from i to j

Ol,Oc - parameters
β-scale parameter
PRi-tour production at zone i
Xij-tour atraction variable l at zone j
Cij-travel cost from zone i to zone j

Calibration of deterrence functions

Deterrence function is an input of the trip distribution model,
however, it can not be observed
Poisson method (initial metrix)
estimation of discrete values for the deterrence
function based on observed (partial) OD-matrix
they lead to determinse function (willingness of
people to travel due to cost and distance))
Hyman method (mean trip length)
estimation of the parameter of an exponential
function based on observed mean trip length,
production and
atraction
OD matrix can be determined by:
using the doubly constrained gravity model
while updating to match the MTL.

1.start with observed productions andattractions,

travel costs and a mean trip length.
Steps:
2.Based on the mean trip length an estimate is
1.Start with F(cij) 1 made for α , with which a doubly constrained
2.Scale to the productions gravity model is applied.
3. Scale to the atractions 3.The resulting mean trip length is checked with
4. Scale the deterrence values such that they the observed value, leading to a new estimate for
represent the OTLD (observed trip length distribution) α . 4.This procedure is repeated until convergence
5. repeat is achieved.

Setting derivatives equal to 0 yields 3 linear equations

in which each parameter is function of the other two
• Similar solution procedure as for trip distribution:
• Determine the constraints
• For each origin i the number of observed trips

Distributing prohibited | Downloaded by Usama Shahdah (usama.elrawy@gmail.com)

 lOMoARcPSD|2396987

(departures)
• For each destination j the number of observed trips
(arrivals)
• For each “bin” k the number of observed trips
• Set all parameters Qi, Xj, Fk
equal to 1
• Determine successively the values for Qi, Xj, Fk until
convergence
Poisson model Hyman’s method
1. Set values for Fk equal to 1 1. Set parameter of the
2. Balance the (partial) matrix deterrence function equal to
for the (partial) production 1/MTL
3. Balance the (partial) matrix 2. Determine values for f(cij)
for the (partial) atraction 3. Balance the matrix for the
4. Balance the (partial) matrix productions and atractions
for the totals per cost class (i.e. apply gravity model)
(i.e. correction in iteration i 4. Determine new estimate for
for estimate of Fk) based on observed MTL and
5. Go to step 2 until computed MTL and go to step
convergence is achieved 2 until convergence is
Hyman’s method Achieved
1. Set parameter of the
deterrence function equal to
1/MTL
2. Determine values for f(cij)
3. Balance the matrix for the
productions and atractions
(i.e. apply gravity model)
4. Determine new estimate for
based on observed MTL and
computed MTL and go to step
2 until convergence is
achieved

Cordon models

Through traffic follows from other source, e.g. license plate survey or other model => through traffic is
thus fixed input and should not be modelled using trip distribution models

Possible approach for cordon models

• Determine production and atraction for internal zones using e.g. regression analysis

• Determine production and atraction for external zones using e.g. counts

• Derive matrix for through traffic (i.e. from cordonzone to cordon zone) from e.g. a regional model

• Subtract through traffic from production and atraction of the external zones

• Apply gravity model with the resulting production and atraction, while making sure that there is no
through traffic, e.g. by setting the travel costs between cordon zones equal to ∞

Distributing prohibited | Downloaded by Usama Shahdah (usama.elrawy@gmail.com)

 lOMoARcPSD|2396987

• Add matrix for through traffic to the resulting matrix of the gravity model

Modal split
Output – how people go, mode choice (bus, car)

Empirical curves Travel time ratio

Data used: observed trips where PT might be atractive

• i.e. train service or ‘express’ service available
• Public transport: Access and egress time to main PT mode, waiting
time first stop, in-vehicle time of main PT mode, waiting time transfer
• Car: travel time based on fixed speeds per area and period of day,
parking

Choice model Derivation nested distribution-modal split model

(use to compare costs between different zones)

Simultaneous distribution- Gravity (entropic) model

modal split using deterrence
functions

Doubly constrained simultaneous distribution/modal split model

Period of day
Output – the most likely time of departure

Time of day – morning, afternoon

Distributing prohibited | Downloaded by Usama Shahdah (usama.elrawy@gmail.com)

 lOMoARcPSD|2396987

Departure time - exactly

A traveler makes a trade-off between:

• A higher travel time due to congestion

• Penalties for deviating from his preferred arrivaltime (PAT)

Departure time choice

Vij - observed utility for trip from i to j at departure time t

Nj -utility of the activity at j

Zij -travel time (without congestion)

Zij(t) -extra travel time due to congestion for departure time t

Ej - penalty for arriving early for departure time t

Lj -penalty for arriving late for departure time t

Point – connection between two links

Link – path between two nodes

Assignment
Out put – what route they take

Flows – how many people choosing each route, travel time, congestion

Distributing prohibited | Downloaded by Usama Shahdah (usama.elrawy@gmail.com)

 lOMoARcPSD|2396987

All or nothing Deterministic use Stochastic Stochastic user

equilibrium (uncongested) equilibrium
(congested)
Choose the For each OD pair,
shortest path all travelers take
their perceived
For each OD pair, shortest route
all travellers take
the shortest route

Uncongested
Shortest Path algorithms

‘tree algorithms’ (Moore and Dijkstra)

For all nodes

• Set travel time t(x) to and set the back node bn(x) to 0

• For the origin i set time to 0 and back node to -1

• Node i is the first active node a

• Select all links (a,j) and check travel times

• If t(a)+time(a,j) < t(j)

t(j)=t(a)+time(a,j) and bn(j)=a and node j becomes an active node

• Node a is no longer active

• Select a new active node from the stack of active nodes and

repeat previous step until there are no active nodes left

• For Dijkstra: select the active node having the lowest travel time

Distributing prohibited | Downloaded by Usama Shahdah (usama.elrawy@gmail.com)

 lOMoARcPSD|2396987

Congested
DUE: Deterministic user equilibrium assignment

Wardrop’s equilibrium law - All travellers choose their optimal route, such that no traveller can improve
his/her travel time by unilaterally changing routes.

Wardrop’s first principle - All used routes have the same travel time which is not greater than the travel
time on any unused route.

Solution principle:

1. Set flows of all links equal to 0

2. Determine link costs based on the link flows (qi

• In first iteration use free flow travel times

3. Perform an assignment (AON) (=intermediate solution wi)

4. Determine link flows

• Naive: Repeat again and again=> Don’t

• Simple: Average flows with previous results (1/n) => MSA

• Smart: Average with optimal step sizes=> Frank-Wolfe

5. Return to step 2

Frank-Wolfe algorithm Main principle:

• Find an initial feasible solution
• Linearise the objective function
• Find a new intermediate
solution
• Determine a new solution in
the direction of the
intermediate solution
• Iterate until the new solution
does not change anymore

SUE: Stochastic user equilibrium assignment

Distributing prohibited | Downloaded by Usama Shahdah (usama.elrawy@gmail.com)

 lOMoARcPSD|2396987

Duality gap in words:

total travel time based on links minus total travel time based on (latest) shortest paths

DUE SUE
Duality gap Zero (wardrop principle) >0

Distributing prohibited | Downloaded by Usama Shahdah (usama.elrawy@gmail.com)

 lOMoARcPSD|2396987

TIGRIS (Transport Infrastructure Land-use (‘Grondgebruik’) Interaction Simulation)

Primarily based on expert judgement

Meant to be a sketch-planning model

TIGRIS XL
Dynamic spatial allocation model

Distributing prohibited | Downloaded by Usama Shahdah (usama.elrawy@gmail.com)