Traffic Flow Theory

& Simulation
S.P. Hoogendoorn

Lecture 1
Introduction
 Photo by Wikipedia / CC BY SA

Course 4821 - Introduction 1 | 57

Photo by wikipedia / CC BY SA
Traffic Flow Theory & Simulation
An Introduction
Prof. Dr. Serge P. Hoogendoorn, Delft University of Technology
2/4/12

Delft
Photo by wikipedia / CC BY SA
University of
Technology

Challenge the future

Introduction
Evacuation Walcheren in case of Flooding

N  120.000 people to evacuate

N  Evacuation time = 6 hour
N  2,5 evacuees / car

N  A58: 2 lanes
N  N57, N254: 1 lane

N  Total available capacity?

  Each lane about 2000 veh/h

  Total capacity 8000 veh/h

N  How to calculate evacuation time

Course 4821 - Introduction 3 | 57

Introduction
Evacuation Walcheren in case of Flooding

N  120.000 people to evacuate

N  Evacuation time = 6 hour
N  2,5 evacuees / car

N  A58: 2 lanes
N  N57, N254: 1 lane
48000
N  Total available capacity?

  Each lane about 2000 veh/h

  Total capacity 8000 veh/h

N  How to calculate evacuation time 8000 veh/h

Course 4821 - Introduction 4 | 57

Introduction
Evacuation Walcheren in case of Flooding

Course 4821 - Introduction 5 | 57

Introduction
Network load and performance degradation

N  It turns out that in simulation only 41.000 people survive!

N  Consider average relation between number of vehicles in network

(accumulation) and performance (number of vehicles completing
their trip)

N  How does average performance (throughput, outflow) relate to

accumulation of vehicles?

N  What would you expect based on analogy with other networks?

  Think of a water pipe system where you increase water pressure

  What happens?

Course 4821 - Introduction 6 | 57

Network traffic flow fundamentals
Coarse model of network dynamics

N  Fundamental relation between network outflow (rate at which trip

end) and accumulation
Network outflow
1. Outflow increases

2. Outflow is constant

3. Outflow reduces

Number of vehicles in network

Course 4821 - Introduction 7 | 57

Network traffic flow fundamentals
Coarse model of network dynamics

N  Fundamental relation between network outflow (rate at which trip

end) and accumulation
Network outflow
1. Outflow increases

2. Outflow is constant

3. Outflow reduces

Number of vehicles in network

Course 4821 - Introduction 8 | 57

Network traffic flow fundamentals
Demand and performance degradation

Course 4821 - Introduction 9 | 57

 Network Performance Deterioration

N  Important characteristic of traffic networks:

  Network production degenerates as number of vehicles surpasses the
critical number of vehicles in the network

  Expressed by the Macroscopic (or Network) Fundamental Diagram

Production
Two (and only two!) causes:
1. Capacity drop phenomenon

Actual
2. Blockages, congestion spillback
production and grid-lock

# vehicles in network
(at a specific time)

Course 4821 - Introduction 10 | 57

Introduction
Lecture overview

N  Traffic queuing phenomena: examples and empirics

N  Modeling traffic congestion in road networks

  Model components of network models

  Modeling principles and paradigms

  Examples and case studies

N  Model application examples

  Traffic State Estimation and Prediction

  Controlling congestion waves

N  Microscopic and macroscopic perspectives!

Course 4821 - Introduction 11 | 57

1.
Traffic Congestion Phenomena
Empirical Features of Traffic Congestion

Course 4821 - Introduction 12 | 57

Historical perspective
Bruce Greenshields

Source: Unknown

Course 4821 - Introduction 13 | 57

First model of traffic congestion
Fundamental diagram

N  Relation between traffic density and traffic speed: u = U(k)

N  Underlying behavioral principles? (density = 1/average distance)

⎛ k ⎞
u = U(k) = u0 ⎜ 1 − ⎟
⎜⎝ k jam ⎟⎠

Course 4821 - Introduction 14 | 57

 Fundamental diagrams
Different representations using q = k×u

()
q=Q k
= kU (k)

u = U (k ) u = U (q)

Course 4821 - Introduction 15 | 57

Dynamic properties
Traffic congestion at bottleneck (on-ramp)

N  Consider bottleneck due to on-ramp

N  Resulting capacity (capacity – ramp flow) is lower than demand
N  Queue occurs upstream of bottleneck and moves upstream as
long as upstream demand > flow in queue (shockwave theory)

36
Driving direction
38 100 9
locatie (km)

40 8
50
7
42 5
0
7 7.5 8 8.5 9 9.5
tijd (u)
Upstream traffic demand

Course 4821 - Introduction 16 | 57

 qupstream C − qon−ramp
Dynamic properties
Shockwave theory

N  Predicting queue dynamics (queuing models, shockwave theory)

N  Predicts dynamics of congestion using FD
N  Flow in queue = C – qon-ramp Q(k 2 ) − Q(k 1 )
N  Shock speed determined by: ω12 =
k −k 2 1

36
Driving direction
38 100 9
locatie (km)

40 8
50
7
42 5
0
Congestion as predicted
by shockwave theory
7 7.5 8 8.5 9 9.5
tijd (u)
Upstream traffic demand

Course 4821 - Introduction 17 | 57

Dynamic features of road congestion
Capacity funnel, instability, wide moving jams

N  Capacity funnel (relaxation) and capacity drop

N  Self-organisation of wide moving jams

Course 4821 - Introduction 18 | 57

 C freeflow
Capacity drop C queue
Two capacities
k
N  Free flow cap > queue-discharge rate
N  Use of (slanted cumulative curves) Pbreakdown
clearly reveals this
N  N(t,x) = #vehicles passing x until t
N  Slope = flow k
q0 = 3700
36
−3800

−3900 38 100 9

−4000 locatie (km)

40 8
50
N(t,x)

−4100 7
42 5
0
−4200 7
7 7.5 8 8.5 9 9.5
−4300 8 tijd (u)
9
−4400 N !(t, x) = N (t, x) " q0 # t
7.5 8 8.5 9 9.5
tijd (u) !
Course 4821 - Introduction 19 | 57
Instability and wide moving jams
Emergence and dynamics of start-stop waves

N  In certain density regimes, traffic is highly unstable

N  So called ‘wide moving jams’ (start-stop waves) self-organize
frequently (1-3 minutes) in these high density regions

Course 4821 - Introduction 20 | 57

Instability and wide moving jams
Emergence and dynamics of start-stop waves

N  Wide moving jams can exist for

hours and travel past bottlenecks
N  Density in wide moving jam is very Start-stop golf met
snelheid 18 km/u
high (jam-density) and speed is low

Gesynchroniseerd Stremming
verkeer

Stremming

Course 4821 - Introduction 21 | 57

Pedestrian flow congestion
Start-stop waves in pedestrian flow

N  Example of Jamarat bridge shows self-organized stop-go waves in

pedestrian traffic flows

Photo by wikipedia / CC BY SA

Course 4821 - Introduction 22 | 57

Pedestrian flow congestion
Start-stop waves in pedestrian flow

N  Another wave example..

Course 4821 - Introduction 23 | 57

2.
Traffic Flow Modeling
Microscopic and macroscopic approaches to
describe flow dynamics

Course 4821 - Introduction 24 | 57

Modeling challenge
Traffic theory: not an exact science!

N  Traffic flow is a result of human decision making and multi-actor

interactions at different behavioral levels (driving, route choice,
departure time choice, etc.)
N  Characteristics behavior (inter- and intra-driver heterogeneity)

  Large diversity between driver and vehicle characteristics

  Intra-driver diversity due to multitude of influencing factors, e.g.
prevailing situation, context, external conditions, mood, emotions
N  The traffic flow theory does not exist (and will probably never
exist): this is not Newtonian Physics or thermodynamics

N  Challenge is to develop theories and models that represent reality

sufficiently accurate for the application at hand

Course 4821 - Introduction 25 | 57

Network Traffic Modeling
Model components and processes

N  Traffic conditions on the 0. Locatiekeuze

Location choice

road are end result of many

decisions made by the 1. Ritkeuze
Trip choice
longer
traveler at different term
2.Destination
Bestemmingskeuze
decision making levels choice

3. Vervoerwijzekeuze
Mode choice
N  Depending on type of
application different levels
4. Routekeuze
Route choice
are in- or excluded in model short
term
demand 5. Vertrektijdstipkeuze
Departure time choice
N  Focus on driving behavior
and flow operations supply
6. Rijgedrag
Driving behavior

Course 4821 - Introduction 26 | 57

Modeling approaches
Microscopic and macroscopic approaches

N  Two dimensions:

  Representation of traffic

  Behavioral rules, flow characteristics

Individual particles Continuum

Individual Microscopic Gas-kinetic models

behavior (simulation) models (Boltzmann equations)

Aggregate Newell model, particle Queuing models

behavior discretization models Macroscopic flow models

Course 4821 - Introduction 27 | 57

Modeling approaches
Microscopic and macroscopic approaches

Helly model:
d
dt
(
v i (t + Tr ) = α ⋅ Δv i (t) + β ⋅ s* (v i (t)) − si (t) )

Individual particles Continuum

Individual Microscopic Gas-kinetic models

behavior (simulation) models (Boltzmann equations)

Aggregate Newell model, particle Queuing models

behavior discretization models Macroscopic flow models

Course 4821 - Introduction 28 | 57

Microscopic models?
Ability to described many flow phenomena

N  Example of advanced micro-simulation model

Photo by Intechopen / CC BY

Course 4821 - Introduction 29 | 57

Modeling approaches
Microscopic and macroscopic approaches

d V(1 / si (t)) − v i (t)

Bando model: v i (t) =
dt τ

Individual particles Continuum

Individual Microscopic Gas-kinetic models

behavior (simulation) models (Boltzmann equations)

Aggregate Newell model, Bando Queuing models

behavior model Macroscopic flow models

Course 4821 - Introduction 30 | 57

Modeling approaches
Microscopic and macroscopic approaches
⎧ ∂k ∂q
⎪ + =r−s
kinematic wave model: ⎨ ∂t ∂x
⎪q = Q(k)
⎩

Individual particles Continuum

Individual Microscopic Gas-kinetic models

behavior (simulation) models (Boltzmann equations)

Aggregate Newell model, particle Queuing models

behavior discretization models Macroscopic flow models

Course 4821 - Introduction 31 | 57

Modeling approaches
Microscopic and macroscopic approaches

∂ρ ∂ ⎛ Ω0 (v) − v ⎞ ⎛ ∂ρ ⎞
∂ρ
Prigogine-Herman model: +v + ⎜ρ ⎟ =⎜ ⎟
∂t ∂x ∂v ⎝ τ ⎠ ⎝ ∂t ⎠ INT

Individual particles Continuum

Individual Microscopic Gas-kinetic models

behavior (simulation) models (Boltzmann equations)

Aggregate Newell model, particle Queuing models

behavior discretization models Macroscopic flow models

Course 4821 - Introduction 32 | 57

3.
Example applications of theory and
models

Course 4821 - Introduction 33 | 57

Application of models
Different models, different applications

N  Ex-ante studies: systematic comparison of alternatives during

different phases of the design process

  Road- and network design

  Traffic Control / Management Strategies and Algorithms

  Impact of traffic information

  Evacuation planning

  Impact of Driver Support Systems
N  Training and decision support for decision makers
N  Traffic state estimation and data fusion
N  Traffic state prediction
N  Model predictive control to optimize network utilization

Course 4821 - Introduction 34 | 57

Application of models
Examples

N  NOMAD model has been extensively

calibrated and validated
N  NOMAD reproduces characteristics
of pedestrian flow (fundamental
diagram, self-organization)
N  Applications of model:

  Assessing LOS transfer stations

  Testing safety in case of emergency
conditions (evacuations)

  Testing alternative designs and
Decision Support Tool

  Hajj strategies and design
NOMAD Animatie by verkeerskunde.nl

Course 4821 - Introduction 35 | 57

Traffic State Estimation & Prediction
Applications of Kalman filters

Photo by fileradar.nl

Course 4821 - Introduction 36 | 57

 Dynamic speed limits
Using Traffic Flow Theory to improve traffic flow

N  Algorithm ‘Specialist’ to suppress

start-stop waves on A12
N  Approach is based on reduced flow
(capacity drop) downstream of wave !

N  Reduce inflow sufficiently by speed-

limits upstream of wave
2
q 1 k
3
1
3
2
k x

Course 4821 - Introduction 37 | 57

4.
Course scope and overview

Course 4821 - Introduction 38 | 57

Scope of course CT4821

N  Operational characteristics of traffic, so not:

  Activity choice and scheduling

  Route choice and destination choice

  Departure time or (preferred) arrival time choice
N  Traffic flow theory does not exclude any transportation mode!
N  Primary focus in this course will be of road traffic (cars) with
occasion side-step to other modes (pedestrian flows)
N  Distinction between

  Macroscopic and microscopic (and something in the middle)

  Flow variables, (descriptive) flow characteristics and analytical tools
(mathematical modelling and simulation)

Course 4821 - Introduction 39 | 57

Overview of flow variables (chapter 2)

Microscopic variables Macroscopic variables

(individual vehicles) (traffic flows)

Flow / volume / intensity

Local Time headway
Local mean speed

Density
Instantaneous Distance headway
Space mean speed

Trajectory Mean path speed

Generalized
Path speed Mean travel time

Course 4821 - Introduction 40 | 57

Flow characteristics

N  Microscopic characteristics (chapter 3)

  Arrival processes

  Headway models / headway distribution models

  Critical gap distributions

N  Macroscopic characteristics (chapters 4 and 5)

  Fundamental diagram

  Shockwaves and non-equilibrium flow properties

Course 4821 - Introduction 41 | 57

Analytical tools

N  Use fundamental knowledge for mathematical / numerical analysis

N  Examples macroscopic tools

  Capacity analysis (chapter 6)

  Deterministic and stochastic queuing models
(chapter 7)

  Shockwave analysis (chapter 8)

  Macroscopic flow models (chapter 9)

  Macroscopic simulation models (13 and 15)
N  Examples microscopic tools

  Car-following models (chapter 11)

  Gap-acceptance models (chapter 12)

  Traffic simulation (chapter 13 and 14)

NOMAD Animatie by verkeerskunde.nl

Course 4821 - Introduction 42 | 57
Course ct4821

N  Lectures by Serge Hoogendoorn and Victor Knoop:

  Monday 8:45 – 10:30

  Tuesday 8:45 – 10:30
N  Mandatory assignment (Hoogendoorn, Knoop + PhD’s):

  Wednesday 13:45 – 17:30 (starting in week 1)

  New Data analysis and FOSIM practicum (week 1-7)

  Reports on assignment

  Description assignment posted on blackboard beginning of next week
N  Course material:

  Parts of the reader (blackboard)

  Assignments (blackboard)

Course 4821 - Introduction 43 | 57

5.
Traffic Flow Variables

Course 4821 - Introduction 44 | 57

 Vehicle trajectories

N  Positions xi(t) along roadway

of vehicle i at time t
N  All microscopic and
macroscopic characteristics
can be determined from
trajectories!
N  In reality, trajectory
information is rarely
available
N  Nevertheless, trajectories
are the most important unit
of analysis is traffic flow
theory

Course 4821 - Introduction 45 | 57

Vehicle trajectories (2)

N  Which are trajectories?

x (a) (b)

(c)

t0 t

Course 4821 - Introduction 46 | 57

Empirical vehicle trajectories

-200
N  Vehicle
trajectories -250
determined using
remote sensing -300
data (from
position (m)
helicopter) at site -350
Everdingen near
Utrecht -400
N  Dots show vehicle
position per 2.5 s -450
N  Unique dataset
340 350 360 370 380 390 400 410
time (s)

Course 4821 - Introduction 47 | 57

Understanding trajectories

d
vi ( t ) = xi ( t )
dt
d2
ai ( t ) = 2 xi ( t )
dt
d3
γ i (t ) = 3 xi (t )
dt

Course 4821 - Introduction 48 | 57

Applications of multiple trajectories

N  Exercise with trajectories by drawing a number of them for

different situations

  Acceleration, deceleration,

  Period of constant speed,

  Stopped vehicles

  Etc.
N  See syllabus and exercises for problem solving using trajectories:

  Tandem problem

  Cargo ship problem
N  Also in reader: discussion on vehicle kinematics described
acceleration ai(t) as a function of the different forces acting upon
the vehicle

Course 4821 - Introduction 49 | 57

 Time headways
N  Time headway hi: passage
time difference rear bumper
vehicle i-1 and i at cross-
section x (easy to measure)

Course 4821 - Introduction 50 | 57

Time headways (3)

N  Headways are local variables (collected at cross-section x0)

N  Mean headway for certain period T of the n vehicles that
have passed x0

1 n T
h = ∑ i =1 hi =
n n

N  Exercise: express the

mean headway for
the cross-section as
a function of the mean
headways H1 and H2
per lane

Course 4821 - Introduction 51 | 57

 Distance headways

N  Distance headway si:

difference between positions
vehicle i-1 and i at time t
(difficult to measure!)

N  Gross distance headway and

net distance headway

Course 4821 - Introduction 52 | 57

Distance headways (2)

N  Distance headway: instantaneous microscopic variable

N  Space-mean distance headway of m vehicles at time instant t for
roadway of length X

1 m X
s= ∑ i m
m i =1
s =

Course 4821 - Introduction 53 | 57

Distance headways (3)

Course 4821 - Introduction 54 | 57

Local, instantaneous and generalized

Generalized measurements Instantaneous measurements

determined for time-space collected at t = 20
region

Local measurements
collected at x = 200

Course 4821 - Introduction 55 | 57

Macroscopic flow variables

N  Traffic intensity q (local variable)

  Vehicle number passing cross-section
x0 per unit time (hour,
15 min, 5 min)

  If n vehicles pass during T,
q is defined by:

n n 1 1
q= = n = =
T ∑ hi 1 n h h
i =1 ∑ i
n i =1
N  Referred to as flow, volume (US)
N  How can flow be measured?

Course 4821 - Introduction 56 | 57

Macroscopic flow variables (2)

N  Traffic density k (instantaneous variable)

  Vehicle number present per unit roadway length (1 km, 1 m) at
instant t

  If m vehicles present on X,
k is defined by

m m 1 1
k= = m = =
X ∑ si 1 m s s
i =1 ∑ i
m i =1
N  Also referred to as concentration
N  How can density be measured?

N  Now how about speeds?

Course 4821 - Introduction 57 | 57

Mean speeds

Course 4821 - Introduction 58 | 57

Mean speeds
N  Local mean speed / time mean speed (see next slide)

  speeds vi of vehicles passing a cross-section x during period T
1 n
uL =
n
∑ v
i =1 i

N  Instantaneous / space mean speed (next slide)

  speed vj of vehicles present at road section at given moment t
1 m
uM = ∑ j =1 v j
m
N  We can show that under special circumstances we can compute
space-mean speeds from local measurements
−1
⎛1 n 1 ⎞
uM = ⎜ ∑i =1 ⎟ (harmonic average local speeds)
⎝n vi ⎠

Course 4821 - Introduction 59 | 57

Homogeneous & stationary variables
N  Consider any variable z(t,x); z is:

  Stationary if z(t,x) = z(x)

  Homogeneous if z(t,x) = z(t)

Course 4821 - Introduction 60 | 57

Stationary flow conditions

Course 4821 - Introduction 61 | 57

Homogeneous flow conditions

Course 4821 - Introduction 62 | 57

Fundamental relation

N  Consider traffic flow that is in stationary and homogeneous

N  Then the so-called fundamental relation holds q = ku
N  Assume intensity q, density k and that all drive with speed u

Number of vehicles
passing cross-section at X
equals the number of
vehicles that is on the
road at t = 0, i.e.

kX = qT

With T=X/u we get

fundamental relation

Course 4821 - Introduction 63 | 57

Which speed to use in q = k×u?

N  Can we apply the fundamental relation q = ku for an

heterogeneous driver population?
N  Yes -> the trick is to divided the traffic stream into homogeneous
groups j of drivers moving at the same speed uj
N  Now, what can we say about the speed that we need to use?

Course 4821 - Introduction 64 | 57

Which speed to use in q = k×u?
N  Can we apply the fundamental relation q = ku for an
heterogeneous driver population?
N  For one group j (vehicles having equals speeds) we have q j = k j u j
N  The total flow q simply equals q = ∑ q j = ∑ k ju j
N  For the total density we have k = ∑ k j
N  Let u = q/k, then
∑ k ju j ⎛ ∑ (q j / u j )u j ∑ qj ⎞
u= = uM ⎜ = = ⎟
∑kj ⎜
⎝ ∑ q j / u j ∑ q j / u j ⎟⎠
N  If we consider qj = 1, then… u = ∑1 = u
∑ j
M
1/ u
N  In sum: the fundamental relation q = kuM may only be
used for space-mean speeds (harmonic mean of individual
vehicle speeds)!!!

Course 4821 - Introduction 65 | 57

Difference arithmetic & space mean

N  Example motorway data time-mean speed and space-mean speed

120

100

80
space mean speed

60

40

20

0
0 20 40 60 80 100 120
time mean speed

Course 4821 - Introduction 66 | 57

What about the derived densities?

N  Suppose we derive densities by k = q/u…

density using space mean speed 200

150

100

50

0
0 50 100 150 200
density using time mean speed

Course 4821 - Introduction 67 | 57

Conclusion average speeds…

N  Care has to be taken when using the fundamental relation q =

kuM that the correct average speed is used
N  Correct speed is to be used, but is not always available from
data!
N  Dutch monitoring system collects average speeds, but which?
N  Same applies to UK and other European countries (except France)

N  Exercise:

  Try to calculate what using the wrong mean speeds means for travel
time computations

Course 4821 - Introduction 68 | 57

Generalized definitions of flow (Edie)

N  Generalized definition flow, mean speed, and density in time-

space plane
N  Consider rectangle T × X
N  Each vehicle i travels distance di
N  Define performance P = ∑ i di
N  P defines ‘total distance traveled’
N  Define generalized flow
P ∑ di / X
q := =
XT T
N  Let X « 1, then di ≤ X
nX n
q= =
XT T

Course 4821 - Introduction 69 | 57

Generalized definition (Edie)

N  Each vehicle j is present in rectangle for some period rj

N  Defined total travel time R = ∑ j rj
N  Generalized definition of density
R ∑ rj / T
k := =
XT X
N  When T « 1, then rj ≤ T
mT m
k= =
XT X
N  Definition of the mean speed
P q ∑ i di
uG = = =
R k ∑ j rj

Course 4821 - Introduction 70 | 57

Overview of variables
Local Instantaneous Generalized
measurements measurements definition (Edie)

Cross-section x Section X Section X

Variable
Period T Time instant t Period T

Flow q (veh/h) q=
n 1
= q = ku q=
∑ d i i
T h XT

Density k (veh/km) k=
q n 1
k= = k=
∑ j j
r
u X s XT

Mean speed u uL =
n
u=
∑ j
vj
u=
q
(km/h) ∑ i (1/ vi ) n k

Course 4821 - Introduction 71 | 57

Furthermore...

N  Homework:

  Read through preface and chapter 1

  Study remainder of chapter 2 (in particular: moving observer, and
observation methods)

Course 4821 - Introduction 72 | 57