Talking traffic: Wireless traffic management

(1)

1

Challenge the future

UMICTA Conference, 9-10 december 2014 University of Stellenbosch, South-Africa

Talking traffic: wireless traffic

management

Prof Dr Bart van Arem

Director TU Delft Transport Institute

2

The Netherlands

World rank South Africa

Area 41.543 km2 135 1.221.037 km2 Population 16.730.623 63 48.601.098 GDP per inhabitant € 34.822 17 $12.100 (large differences) x30 x3 South Africa Capital Pretoria 3

Infrastructure of the Netherlands

Waterway density Liechtenstein 0,18 Netherlands 0,15 Belgium 0,07 Bangladesh 0,06 Vietnam 0,05 Roadway density Belgium 5,03 Netherlands 3,29 Japan 3,20 Hungary 2,12 Jamaica 2,01 Railway Density

Czech Republic, Switzerland, Germany 0,12

Belgium 0,11

Hungary 0,09

Cuba, Austria 0,08 Netherlands, Slovakia, Japan, United Kingdom 0,07

South Africa 0,44 (0,05 paved) South Africa 0.02

4

Mainports of the Netherlands

1

st

in Europe

10

th

_{world wide}

Largest non Asian port

3% Dutch GDP

Quality:

1

st

in Europe

4

th

world wide

Europe

4

th

_{passenger volume}

3

rd

freight volume

4,7% Dutch GDP (incl

airlines)

Port of Rotterdam

Schiphol airport

Traffic management in the

Netherlands

• Motorway monitoring and signalling

• Ramp metering

• Dedicated lanes

• Operation and maintenance 1B€ py

Motorways with daily volume over 180.000 vehicles Netherlands 15 United Kingdom 2 Germany 2

Shifting from

road-side to

in-vehicle

paradigms

Talking

traffic

(2)

7

Can we solve or mitigate

congestion using cooperative

in-vehicle systems?

8

Theoretical principles

1. Prevent spill-back of

queues 2. Increase throughput

3. Manage inflow into (sub-) network

4. Distribute traffic over network efficiently

9

Typical in-vehicle systems

ACC

predecessor

Lane, speed,

headway advices

Route navigation

>2 km

Tactical

driving

advice

Social

navigation

10

Tactical driving advice

• In-vehicle advice to the driver on speed, headway and lane use

• Comfort, safety but also efficiency

• Fusion of in-car and road side traffic data

• Perspective on large scale implementation

• Retrofit

• Integrating existing technologies

Integrating geographical levels

direct surroundings x kilometre ahead (e.g. 2 km) regional network vehicle dynamics • Intelligent camera

• eHorizon • CCC Service Centre • Traffic Management Centre • enhanced positioning • GNSS • (vehicle sensor data) • speed advice based on on-coming road geometry and speed limits • current speed limit from traffic sign, • speed advice based on local traffic situation • speed, headway and lane use advice • dynamic maximum

speed

CCC integrated speed, headway and lane use advice CCC-vehicle

driving direction

Traffic flow improvement

algorithm

Traffic Management Centre

Floating Car Data

Loop detector data

Traffic state estimation & prediction

Advice algorithm

(3)

13

CCC Implementation

Embedding privacy,

security and fault

tolerance

14

Lane level traffic state prediction

time detector data delay (~65s)

shockwave trajectory equipped vehicle detector detector detector

• Based on Adaptive Smoothing

Method (ASM)

• Propagates traffic state according

to typical speeds

• Can be used for short-term

predictions

• 1 minute prediction to allow

drivers time to act

• Used at lane level

15

Advice principles

• Distribution

Redistribute traffic over lanes to prevent breakdown

• Acceleration

Decrease capacity drop at end of congestion

• Spillback

Prevent spill back

lmerge lbusy ladj lmerge lbusy ladj lbusy ladj lmerge lbusy ladj lmerge lbusy ladj B A A A B B a) b) c)

Congestion Free flow

Lane change/keep advice Yield advice

Speed/synchronize advice Short/safe headway advice

Trigger

A/B Applicable destination

16

Test drives

December 2012

1 equipped vehicle

14 subjects, 3-4 test drives per subject

12 km stretch

start

end

Potential results using simulation

40-50% delay reduction at 100% penetration and compliance

0 10 2030 40 50 60 70 8090 100 180 200 220 240 260 280 300 320 340 360 380 ω (compliance) = λ (penetration) Me an tra ve l tim e d ela y [s] All Acceleration Distribution 0 10 20 30405060 708090100 200 250 300 350 400 450 500 550 ω (compliance) = λ (penetration) M ean t ra vel ti m e d ela y [s ] All Acceleration Distribution Spillback 0 0.10.20.30.40.50.60.70.80.9 1 0 10 20 30 40 50 60 70 80 90 ω (compliance) = λ (penetration) C ong estion d u ra tion [m in ] 1800 1900 2000 2100 2200 2300 2400 2500 Flo w [veh /h]

Congestion lane drop Congestion Moordrecht Saturation flow Max. lane intensity Max. agg. intensity

0 0.10.20.30.40.5 0.60.70.80.9 1 20 25 30 35 40 45 50 55 60 65 70 ω (compliance) = λ (penetration) C onge stion d ura tion [ m in] 2200 2300 2400 2500 2600 2700 2800 2900 3000 3100 F low [ve h/h ]

Congestion lane drop Congestion Moordrecht Saturation flow Max. lane intensity Max. agg. intensity

(4)

19

Findings

• Tactical driving advice works technically and functionally

• Add-on to existing connected navigation

• Improved localization and timing needed

• Need for coordinated CCC network wide

20

Typical in-vehicle systems

ACC

predecessor

Lane, speed,

headway advices

Route navigation

>2 km

Tactical

driving

advice

Social

navigation

21

Always the fastest route

Traffic data centre

Position, speed

Travel times

Inductive

loop data

Tracking of

cellular phones

GPRS

Is the fastest route also the best route?

22

User equilibrium and system

optimum

Consider route choice in a network

User equilibrium: each driver has chosen a route that

individually optimises own travel time

System optimum: route chosen such that total travel time in

the network is optimised

Individual route guidance reinforces user equilibrium

orientation

Total travel time in user equilibrium 30-40% higher than in

system optimum

Selfish or social?

α

=0.2,

β

=4 and C=1000

τ

=10, q=900

τ

=15, q=400

current

situation

fastest route

social route

q

900

400

901

400

900

401 T

11,31

15,08

11,32

15,08

11,31

15,08

TT

10181

6031

10198

6031

10181

6046

STT

16212

16228

16227

Social navigation

Personal

cost

System

cost

Social

cost

=

+

Altruism

level

*

Altruism factor x: 1 system unit equivalent to x personal units

x=0.25: sacrifice 1 minute personal time to reduce 4 minutes system time

(5)

25

Modelling framework

• Simulation with time step 1s

• Longitudinal behaviour (Gipps, 1981)

• Lane change behaviour (Gipps, 1986)

• speed advantage

• mandatory,

• gap searching

• gap acceptance

26

‘ Bay area’ network

27

UMICTA Conference, 9-10 december 2014

University of Stellenbosch, South-Africa Challenge the future28