35
WWW.PTV.DE Estimating Real-Time Urban Traffic States in VISUM Online (PTV TrafficPlatform) PIARC Conference – Kuala Lumpur 2006 Dr. Gerhard Ploss, Dr. Peter Vortisch
WWW.PTV.DEWWW.PTV.DE
Estimating Real-Time Urban Traffic Statesin VISUM Online (PTV TrafficPlatform)
PIARC Conference – Kuala Lumpur 2006Dr. Gerhard Ploss, Dr. Peter Vortisch
2© PTV AG 2006
PTV TrafficPlatform
Back
No „Big Brother“but „Intelligent Traffic Management“....
3© PTV AG 2006
PTV TrafficPlatform
Back
... Including the Needs of New Mobile Services...
„Maybe you are right,Darling... But I followexactly my new navigation system...“
4© PTV AG 2006
PTV TrafficPlatform
Back
Traffic State as Basis for Information and ControlTraffic state contains> for the road user:
Congestion or free flow
> for navigation devices:Travel times
> for variable direction signs and traffic forecast:Traffic values, ODs and routes
A current traffic state requires> current measurement data
> fast processing
5© PTV AG 2006
PTV TrafficPlatform
Back
PTV TrafficPlatform (VISUM Online) – Objectives
•Data management•Data distribution•Estimation of traffic state•Forecast & Scenarios•Incident management
PTV TrafficPlatformInput
•Detector data•Floating car data •Traffic reports •Road-works•Networks•OD-Matrices
Output
•Visualization •Mobile services•Dynamic route guidance•Evaluation & Reports•Transportation planning•Traffic control
6© PTV AG 2006
PTV TrafficPlatform
Back
Intelligence
•OD Estimation (Calibration) and Routing
•Algorithms for Urban Traffic (Propagation)
•Algorithms for Motorway Traffic (ASDA/FOTO)
•Floating Car Data
•Data Fusion
•Forecast & Scenarios
•Cluster Analysis
•Statistics & Evaluation
7© PTV AG 2006
PTV TrafficPlatform
Back
Intelligent Models consider Road Networks...
Measurements Local Models Network Models
8© PTV AG 2006
PTV TrafficPlatform
Back
Estimation of Traffic State and Forecast
Data completition
15 min
Short-term forecast
15 min
Strategy A
Strategy B
9© PTV AG 2006
PTV TrafficPlatform
Back
Work Flow of the Traffic State Estimation
Historicdetector values
hourly basedOD-Matrices
Trafficmessages
Current routes
Route estimation (Assignment)
Current detector values
Measurementpropagation
Network model
Cluster-analysis
typical day pattern
Base-OD-Matrices
Matrixcalibration
current traffic state
offline
online(every 5 minutes)
Time interval1 hour
Time interval5 minutes
10© PTV AG 2006
PTV TrafficPlatform
Back
Flow Bundle
Detektordetector
11© PTV AG 2006
PTV TrafficPlatform
Back
Innerurban Traffic: Measurement Propagation
Propagation of measured traffic volumes along routes in the network
80 %
20 %
50 %
50 % M
M
12© PTV AG 2006
PTV TrafficPlatform
Back
Requirement: Sensitivity to Real-time Events
Impact on link: Resulting impact on network:
13© PTV AG 2006
PTV TrafficPlatform
Back
Local Forecast by Time Series Selection
Idea: Traffic patterns for comparable days are similar, because the activities of mobile persons are repetitive.
Task: Recognition of the correct pattern for the day
now
14© PTV AG 2006
PTV TrafficPlatform
Back
Short-term Forecast
t
t
t
t
t t
t
t
t
detection until t0 detection until t0+ ∆t detection until t0+ 2∆t
V V V
state t0- ∆t state t0 state t0+ ∆t state t0+ 2∆t
Idea: Forecasting of local measurement values and use of the propagation method
15© PTV AG 2006
PTV TrafficPlatform
Back
Model-based Data Completion
time
Spatial interpolation:
Temporal extrapolation: Forecasting
simple, but wrongbetter: model based
16© PTV AG 2006
PTV TrafficPlatform
Back
Model for Freeway Traffic: ASDA/FOTO by Prof. Kerner
v (km/h)
Zeit Ort (km) Ort (km)
synchronisierter Verkehr
sich bewegen-der langer Stau
q (Fzg./h)
Kerner, Journal of Physics A, 2000
v (km/h)
Zeit Ort (km) Ort (km)
synchronisierter Verkehr
sich bewegen-der langer Stau
q (Fzg./h)
Kerner, Journal of Physics A, 2000
Theory on physics of traffic flow: 3 phases (free, „synchronised“, jam)
synchonized traffic
wide moving jamtime space (km)
q (veh/h)
17© PTV AG 2006
PTV TrafficPlatform
Back
ASDA-FOTO with Sparse Detection
x [km]
time8:00 8:30 9:007:30
470
475
480
485
490
9:30
(a)
loca
tions
of d
etec
tors
use
d
x [km]
time8:00 8:30 9:007:30
470
475
480
485
490
9:30
(b)
loca
tions
of d
etec
tors
use
d
Weg-Zeit-Diagramme A5-Süd: (a) 31 Detektoren; (b) 9 der 31 Detektoren
x [km]
time8:00 8:30 9:007:30
470
475
480
485
490
9:30
(a)
loca
tions
of d
etec
tors
use
d
x [km]
time8:00 8:30 9:007:30
470
475
480
485
490
9:30
(b)
loca
tions
of d
etec
tors
use
d
Weg-Zeit-Diagramme A5-Süd: (a) 31 Detektoren; (b) 9 der 31 Detektorenspace-time-diagrams A5 southbound: (a) 31 detectors; (b) 9 detectors
18© PTV AG 2006
PTV TrafficPlatform
Back
Data Source: Taxi Cab - FCD
Positioning of taxi cabs via GPS
Position is sent to operating center viataxi channel (no additional cost!)
High frequency ofposition reports(less than 2 minutes)
Routing between reported positionsin the operation center
Source: DLR
19© PTV AG 2006
PTV TrafficPlatform
Back
Filtering of Exclusive Lanes for Taxi Cabs and Buses
20© PTV AG 2006
PTV TrafficPlatform
Back
Data Source: FCD from Buses
Position reportedevery 2 minutes
Map-matching considers line routes
Dwell times at stops considered
21© PTV AG 2006
PTV TrafficPlatform
Back
Comparison of Speeds from FCD vs. Local Detection
0
20
40
60
80
100
120
140
160
0 20 40 60 80 100 120 140 160 180
V_FCD[km/h]
V_D
ET [k
m/h
]
FCD
localdetection
freeway
22© PTV AG 2006
PTV TrafficPlatform
Back
System Architecture
Measurements
Incidents
OD matrices
Networks
Input
Probes/FCD
Output
TP exportEx
porter
Traffic situation
Automated reports
Calibrated matrices
Up-to-date networks
ForecastTP
importTP
Database
TP Network Editor (VISUM)
TMFUser Interface
Aggregation
Traffic State
Forecast
Calculation
Evaluation
VISUM
IncidentsTraffic situation & Forecast
OD Matrices
Measurement
Network
23© PTV AG 2006
PTV TrafficPlatform
Back
TP Database> Standard for
Traffic Management> Relational database for
> ORACLE> MS SQL Server> MS ACCESS> mySQL
> 300 tables in numerous modules
Static
NetworkDetectorsPOIs
Dynamic
Raw dataAggregated dataResults
24© PTV AG 2006
PTV TrafficPlatform
Back
Overview - PTV TrafficManagementFramework (TMF)
> Graphical User Interface for Traffic Management
> Main Tasks
> Visualization
> Evaluation
> Management
> Access to content of a database> Intuitive operation> Flexible customization> Desktop and browser solution
25© PTV AG 2006
PTV TrafficPlatform
Back
TMF - Overview Framework
Display window
Filter
Navigation Smart map
Layercontrol
Furtherapplication
Detailedinformation
26© PTV AG 2006
PTV TrafficPlatform
Back
Incident Manager
> Different sources> Different categories
> Unique referencing> Solution of conflicts> Support of TMC, TPEG and other formats
> Defines impact on traffic:(traffic condition and dynamic route guidance)> Location: Number of links in the network> Time: Valid period of an incident> Effect: Reduction of capacity and speed
27© PTV AG 2006
PTV TrafficPlatform
Back
Browser Solution
28© PTV AG 2006
PTV TrafficPlatform
Back
What is the VMZ Berlin?
> Actually, the VMZ Berlin (Verkehrsmanagementzentrale) mainly is a Traffic Information Center
> An interface to the VKRZ Berlin(Traffic Control Center) is in progress.
> Contracted in 1999 by the Senat of Berlin;in operation since 1.7.2003
> Operated by: VMZ Berlin(Siemens and DaimlerChrysler)
29© PTV AG 2006
PTV TrafficPlatform
Back
What the VMZ Berlin provides
„Comprehensive and up-to-date mobility servicesfor all transport users and all means of transport in Berlin“
Better traffic information shall provide the basis for a moreefficient traffic flow.
Information broadcast media:> Internet
> Variable message signs on roads
> mobile phone services (SMS, WAP)
> radio stations
30© PTV AG 2006
PTV TrafficPlatform
Back
Traffic Management Centre Berlin
>ca. 400 Detectors>Network with ca. 10.000 links>OD Matrix with ca. 1000 zones>Traffic State on ca. 3000 links
31© PTV AG 2006
PTV TrafficPlatform
Back
www.vmzberlin.de
32© PTV AG 2006
PTV TrafficPlatform
Back
Quality Check of the Estimated LOS
Laboratory experimentsScenario as part of approvalExamination with additional detection> Mobile detection devices> 14 locations
per day> 10 15-min-periods relevant> Considered: flow and speed => LOS> Compared: calculated and predicted
LOS to measured situation
33© PTV AG 2006
PTV TrafficPlatform
Back
Comparison of VolumesB1: Invalidenstraße
0
200
400
600
800
1000
1200
1400
1600
1800
2000
2200
06:1
506
:30
06:4
507
:00
07:1
507
:30
07:4
508
:00
08:1
508
:30
08:4
509
:00
09:1
509
:30
09:4
510
:00
10:1
510
:30
10:4
511
:00
11:1
511
:30
11:4
512
:00
12:1
512
:30
12:4
513
:00
13:1
513
:30
13:4
514
:00
14:1
514
:30
14:4
515
:00
15:1
515
:30
15:4
516
:00
16:1
516
:30
16:4
517
:00
17:1
517
:30
17:4
5
[Fz/
h]
B3: Treskow-Allee
0
200
400
600
800
1000
1200
1400
1600
1800
2000
2200
06:1
506
:30
06:4
507
:00
07:1
507
:30
07:4
508
:00
08:1
508
:30
08:4
509
:00
09:1
509
:30
09:4
510
:00
10:1
510
:30
10:4
511
:00
11:1
511
:30
11:4
512
:00
12:1
512
:30
12:4
513
:00
13:1
513
:30
13:4
514
:00
14:1
514
:30
14:4
515
:00
15:1
515
:30
15:4
516
:00
16:1
516
:30
16:4
517
:00
17:1
517
:30
17:4
5
[Fz/
h]
B4: Heerstraße
0
200
400
600
800
1000
1200
1400
1600
1800
2000
2200
06:1
506
:30
06:4
507
:00
07:1
507
:30
07:4
508
:00
08:1
508
:30
08:4
509
:00
09:1
509
:30
09:4
510
:00
10:1
510
:30
10:4
511
:00
11:1
511
:30
11:4
512
:00
12:1
512
:30
12:4
513
:00
13:1
513
:30
13:4
514
:00
14:1
514
:30
14:4
515
:00
15:1
515
:30
15:4
516
:00
16:1
516
:30
16:4
517
:00
17:1
517
:30
17:4
5
[Fz/
h]
B7: Warschauer Straße
0
200
400
600
800
1000
1200
1400
1600
1800
2000
2200
06:1
506
:30
06:4
507
:00
07:1
507
:30
07:4
508
:00
08:1
508
:30
08:4
509
:00
09:1
509
:30
09:4
510
:00
10:1
510
:30
10:4
511
:00
11:1
511
:30
11:4
512
:00
12:1
512
:30
12:4
513
:00
13:1
513
:30
13:4
514
:00
14:1
514
:30
14:4
515
:00
15:1
515
:30
15:4
516
:00
16:1
516
:30
16:4
517
:00
17:1
517
:30
17:4
5
[Fz/
h]
34© PTV AG 2006
PTV TrafficPlatform
Back
Results from Field Test
77 72 80 73
17 23 14 21
6 5 6 6
0%10%20%30%40%50%60%70%80%90%
100%
currentstate testintervals
forecast test
intervals
currentstate whole
day
forecast whole day
LOS wrongoff by oneLOS correct
35© PTV AG 2006
PTV TrafficPlatform
Back
PTV Planung Transport Verkehr AG, 76131 Karlsruhe
WWW.PTV.DE
