Title: TRAFFIC PERFORMANCE MODELS I: TRAFFIC FLOW THEORY AND SIMULATION APPROACHES
1TRAFFIC PERFORMANCE MODELS I TRAFFIC FLOW
THEORY AND SIMULATION APPROACHES
- Haris N. Koutsopoulos
- Northeastern University
- M.I.T. Summer Professional Program 1.10s
- Modeling and Simulation for Dynamic
Transportation Management Systems - July/August 2003
-
2Outline
- Basic Elements
- Macroscopic models
- Traffic Stream Models
- Continuum Models
- Microscopic models
- Car Following Models
- Lane Changing
- Traffic Simulation Models
3Basic Elements of Traffic Flow Theory
- flow (q) number of vehicles crossing a point
per unit of time - speed (u)
- time-mean speed ut average speed of vehicles
crossing a point of the roadway - space-mean speed us average speed of vehicles
over a roadway segment at any given time, or
based on average time it takes to cross the
segment - concentration
- density (k) number of vehicles per unit length
of roadway - occupancy (o) percent of time a point of the
road is occupied - headway
- space headway (s) distance between two
consecutive vehicles - time headway (h) time between passage of two
consecutive vehicles
4Measurements
- Fixed Point
- Fixed Time (aerial photograph)
- Moving observer
5Time-Space Diagram Analysis at a Fixed Position
6Time-Space Diagram Analysis at a Fixed Time
7Measurements
8Measurements
At a fixed location, time T ui instantaneous
speed
Observation at fixed location over time T
Instantaneous photograph, speeds over section of
length L
- time-mean speed ? space-mean speed (typically
ut 1.06 1.12us)
9Fundamental Relationship
- q us ? k
- where,
- q flow
- us space-mean speed
- k density
10Traffic Stream Models
- Objective
- Provide fundamental relationships among
macroscopic traffic stream characteristics for
uninterrupted flow conditions - speed-density
- flow-density
- speed-flow
11Field Data (Holland Tunnel, Eddie 63)
12Relationships
- Structure
- Single regime same functional form under all
traffic conditions - Multiple regime different models for different
traffic conditions
13Examples
uf, kj, parameters to be calibrated
14Examples
- Greenbergs model ?
- Underwoods model ?
uf, kj, km, parameters to be calibrated
15Examples
- Edies model
- Uncongested region
- Underwoods model
- Congested region
- Greenbergs model
- Discontinuity
16Speed-Density Relationships
17Flow-Density Relationships
18Speed-Flow Relationships
19Empirical Results
20Speed-Flow Relationship
Source HCM 2000
21Continuum Flow (Kinetic) Models
- density at time t k
- density at time tdt kdk
- conservation of flow qdt kdx (q dq)?dt
(kdk)?dx
22Continuum Flow (Kinetic) Models, contd
- First order models (Lighthill and Whitham,
Richards, LWR)
Conservation of flow
Basic equation of flow
Equilibrium speed (e.g. Greenshields
speed-density model)
q flow u space-mean speed k density g(x,t) gen
eration rate
23Solution of First Order Models
- Problem
- Given a set of initial and boundary conditions,
- e. g. k0 k(x0, 0)
- Find k(x,t)
- Solution
- Exact solution is based on the method of
characteristics (waves) lines with the same
density
24Solution of First Order Models
- Properties
- characteristics (waves) are straight lines along
which the density is constant (equal to the
density at the initial point they emanate from) - a disturbance at a point propagates along the
characteristic through that point - a disturbance travels at a speed given by the
slope of the characteristic
25Solution of First Order Models
26Comments
27Comments
- If (A, B) (2, 1) w21 is positive (shockwave
moves forward) - If (A, B) (3, 2) w32 is negative (shockwave
moves backward)
28Example Traffic Lights
time
29Shockwaves at Traffic Lights (red)
Stopping waves
uw
A
qA
Flow (q)
B
D
qD 0
kj
kA
Density (k)
30Shockwaves at Traffic Lights (green)
Starting waves
uw
C
qB qmax
Flow (q)
B
D
kj
kC
31Discussion on First Order Models
- Issues
- driver differences
- motion through a shock
- stability
- Implications
- light traffic
- desired speed
- platoon formation
- restricted traffic
- instantaneous change of speed
- stop and go instabilities
32High Order Continuum Flow Models
- Attempt to represent the acceleration of the
traffic stream - Acceleration f(traffic ahead, equilibrium
speed, reaction time,)
33High Order Continuum Flow Models
conservation of flow
basic equation of flow
dynamic speed equation
- where,
- T reaction time, relaxation parameter
- ? anticipation parameter
- f(k) equilibrium speed (speed-density model)
- Rate of change of speed is a function of
- Convection (tendency to travel upstream with
current speed) - Relaxation to equilibrium
- Anticipation (impact off density ahead)
34Microscopic Models
- Car-following
- Lane changing
- Gap acceptance
35Car-Following Models
Common Model Response(t) sensitivity?stimulus(t
-T) T reaction time Stimulus ?v (relative
speed), ?x (relative distance) Sensitivity
function of ?x, speed, traffic conditions Response
acceleration, speed
36Car-Following Models
GM models (Herman, Gazis)
acceleration of vehicle n at
time t ?, l, m parameters
37Car-Following Models
- Distance-based models (Newell)
- Stimulus is a function of relative distance
- Psycho-physical spacing model (Wiedemann, 1974)
- reaction based on relative speed-relative
distance combinations - General acceleration models
- Ahmed, Ben-Akiva, Koutsopoulos (2002)
- Multiple regimes of traffic
- Free flowing
- Car-following
- Time headway threshold and reaction time
distributions
38General Acceleration, contd
- Ahmed, Ben-Akiva, Koutsopoulos (2002)
- Car following
- Distributed headway threshold and reaction time
39Car-following and Traffic Stream Models
mgt
- Are car-following and traffic stream models
consistent? - Under steady-state, single lane conditions
integration of car-following models results in
traffic stream models
40Flow Models Derived from Car-Following Models
41Lane Changing Models
- Mandatory and discretionary lane-changing
- mandatory getting off the current lane in order
to continue on the desired path (e.g. exiting),
or to avoid lane closure - discretionary attempting to achieve desired
speed, avoid following trucks, avoid merging
traffic, etc.
42Lane-changing, contd
- Ahmed, Ben-Akiva, Koutsopoulos (1999)
- Mandatory and discretionary
- 3 levels
- Decision to consider lane change
- Target lane choice for DLC
- Random utility framework
- Gap acceptance
- Lead and lag gaps
- Forced merging
43Dimensions of Lane Changing
44Gap Acceptance
- Critical Gap
- if available gap lt critical gap reject the gap
- if available gap gt critical gap accept the gap
- Lead and lag gap
- Critical gap is a function of
- Relative speed
- First gap
- Number of gaps rejected
- Remaining length (mandatory lane changing)
- Other opportunities
- Traffic conditions
The Critical Gap is Function of explanatory
variables Gng (t) expXng (t)?g ?ng (t)
45Limitations of Existing Models
- Independent behaviors
- Acceleration and lane-changing
- Mandatory and discretionary lane changing
- Reactive
- No anticipation
- Myopic
- No planning
46Other Traffic Performance Models
- Delay models
- Queuing theory (deterministic and stochastic
approaches)
- Empirical models
- Simulation models
- Synthesis of traffic theory models, delay models,
traffic dynamics representation
47Simulation Models
- Definition
- the process of designing a model of a real
system and conducting experiments with this model
for the purpose either of understanding the
behavior of the system or of evaluating various
strategies (within the limits imposed by a
criterion or set of criteria) for the operation
of the system (Shannon, 1975) - Approaches
- Discrete event
- synchronous
- Asynchronous
- Continuous time
- Hybrid
48Simulation Models
- Functionality and level of detail
- Network representation
- Flow representation
- Traffic dynamics
- Support of control strategies
- Surveillance
- Travel behavior/demand
- Overall structure
- Event-based
- Time-based
- Output
- Measures of effectiveness (MOEs)
49Level of Detail
- Based on their flow and traffic dynamics
representation traffic simulation models are
characterized as - Macroscopic
- Fluid representation of flow
- Time and space discretization
- Mesoscopic
- Individual vehicle representation
- Continuous space
- Usually discrete time
- Microscopic
- Individual vehicle representation
- Traffic dynamics through vehicle interactions and
movements - Nanoscopic
- Many common elements with microscopic
- Detailed representation of vehicle dynamics
50Overall StructureTime- vs. Event-Based Simulation
- Time-based models advance the clock at fixed
intervals ?t - ?t may be different for different processes
- choice of ?t important for
- Efficiency
- Accuracy
- Event-based models maintain an event list. First
event in the list is processed next - sequencing of events may be difficult
- less control over efficiency
51Macroscopic Model Characteristics
- In effect numerical solutions to continuum flow
models - Usually deterministic
- Common for evaluation of freeway corridor
operations - Traffic dynamics
- queuing theory
- kinetic theory
- simple input/output
- simple continuum
- high order continuum
- Basic approach
- numerical solution of kinetic equations
- discretization of space and time
52General Approach
- define a grid of points (n, j) in the time-space
domain - replace derivatives by finite differences
- Forward, Backward, Central
- Lax-Friedrichs scheme
- Codunovs finite difference approximation
53Macroscopic Model Example
- Numerical solution based on Laxs method
?j coefficient for converting flow to density
(e.g. ?t/?xj)
54Paynes model (High Order)
55Macroscopic Models Examples
56NETCELL (Daganzo)
- Characteristics
- Dynamic evolution of multi-commodity traffic over
large scale freeway networks - Links divided into cells traversed in one
simulation step (under free flow speed) - Consistent with hydrodynamic theory of flow (cell
transmission model) - Time-based, with an event list
- Inputs
- network geometry
- incidents
- O-D flows
- routing information
57NETCELL, contd
58NETCELL, contd
cell i1
cell i-1
ni(t1) ni(t) yi(t) - yi1(t) yi(t) min
ni-1(t), Qi(t), Ni(t) - ni(t)
- l length of cell
- ni(t) number of vehicles in cell i at time t
- Ni(t) max number of vehicles in cell i
- Qi(t) capacity flow into i for time interval t,
and - yi(t) number of vehicles that flow from cell i-1
to cell i in (t, t1) - Note results are independent of the processing
order of cells
59Macroscopic Models Comments
- Debate first order vs. high order models
- theoretical considerations
- empirical evidence
- Recent developments include
- effects due to lane drops
- freeway-to-freeway interactions
- bus effects
- Importance of method for numerical solution
- Numerical solution requires appropriate
discretization of time and space
60Macroscopic Models Comments
- Advantages
- data requirements
- execution speed
- Usual output aggregate measures of performance
- queue lengths
- speed contours
- total delay
- total travel time
- fuel/pollution statistics
61Mesoscopic Traffic Simulation Models
- Flow Representation
- individual vehicles or groups of vehicles with
similar characteristics (packets) - Traffic Dynamics
- fluid approximation
- queuing theory
- Network Representation
- integrated networks or corridors
- link-based
- lane-based
- Traffic Control
- aggregate by equivalent capacities
- Detailed
- Structure
- time-based
62Mesoscopic Simulation Models Examples
- Microscopic version has also been developed
63Microscopic Traffic Simulation Models
- Detailed
- Synthesis of models
- Driving behavior
- Travel behavior
- Control and routing strategies
- Driver classes
- Integrated networks (freeways, urban streets)
- Usually time-based (e.g. 0.1 sec. time step)
- Stochastic
64Microscopic Traffic SimulationApplication Needs
65Microscopic Traffic SimulationFunctionality
Needs
66Examples of Microscopic Simulation Models
- AIMSUN 2 Universitat Politècnica de Catalunya,
Spain - ANATOLL ISIS and Centre dEtudes Techniques de
lEquipement France - ARTEMiS University of New South Wales, School of
Civil Engineering Australia - ARTIST Bosch Germany
- CASIMIR Institut National de Recherche sur
les Transports et la Sécurité France - CORSIM Federal Highway Administration
USA - DRACULA Institute for Transport Studies,
University of Leeds UK - FLEXSYT II Ministry of Transport
Netherlands - FREEVU University of Waterloo, Department
of Civil Engineering Canada - FRESIM Federal Highway Administration
USA - HUTSIM Helsinki University of
Technology Finland - INTEGRATION Queens University, Transportation
Research Group Canada - MELROSE Mitsubishi Electric Corporation
Japan - MICROSIM Centre of parallel computing (ZPR),
University of Cologne Germany - MICSTRAN National Research Institute of Police
Science Japan - MITSIMLab Massachusetts Institute of
Technology USA - NEMIS Mizar Automazione, Turin
Italy
67Examples of Microscopic Simulation Models, contd
- PADSIM Nottingham Trent University NTU UK
- PARAMICS The Edinburgh Parallel Computing Centre
and Quadstone UK - PHAROS Institute for simulation and
training USA - PLANSIM-T Centre of parallel computing (ZPR),
University of Cologne Germany - SIGSIM University of Newcastle UK
- SIMDAC ONERA Centre d'Etudes et de Recherche
de Toulouse France - SIMNET Technical University Berlin
Germany - SISTM Transport Research Laboratory,
Crowthorne UK - SITRA-B ONERA Centre d'Etudes et de Recherche
de Toulouse France - TRANSIMS Los Alamos National Laboratory USA
- THOREAU The MITRE Corporation USA
- VISSIM PTV System Software and Consulting GMBH
Germany
A detailed description of microscopic models can
be found in www.its.leeds.ac.uk/smartest
68TRANSIMS Microsimulation
intersection with multipleturn buffers (not
internallydivided into grid cells)
single-cell vehicle
multiple-cell vehicle
7.5 meter ? 1 lane cellularautomaton grid cells
69TRANSIMS Microsimulation, contd
- For all particles i simultaneously
- IF ( vi gt gapi ? deceleration)
- ELSE IF (vi lt vmax ? acceleration)
- ELSE (vi vmax and vi lt gapi ? free-flow)
70Nanoscopic Simulation Models
- General characteristics
- detailed driver behavior
- detailed vehicle dynamics model
- detailed vehicle-drive interactions
71Applications of Traffic Simulation Models
- Evaluation at the planning and policy level
- Evaluation at the operational level
- traffic control
- intersection/urban street operations
- freeway corridors
- ITS
- Automated Highway Systems (AHS)
- Public transportation
- Design
- Optimization
72Applications of Traffic Simulation Models
- Real time decision support systems
- Prediction
- Strategy design/evaluation
- route guidance
- traffic control
- Research and development
- New concepts and algorithms
- Role
- used independently
- elements of larger systems
- supply representation
- dynamic network loading models (DNL)
73Applications of Traffic Simulation Models
74Summary
- Integrated networks
- Traffic dynamics
- queue build-up and spillbacks
- Travel behavior and demand
- Driver characteristics and vehicle classes
- Dynamic Traffic Management
- Trends
- Hybrid models
75General Comments
- Model complexity
- inputs
- Calibration
- Validation
- program is a close approximation to reality
- Estimation and calibration
- Output analysis
- large amounts of output
- interpretation of results
- Sources of error
- functional
- distributional
- independence
- aggregation
- boundary effects