transport thursday 20 september 2012 - why we need a virtual traffic and travel laboratory to...

53
Challenge the future 1 Transport Thursdays An initiative of the TU Delft Transport Institute Dr. ir. Hans van Lint (CEG/T&P) Why we need a virtual Traffic and Travel Laboratory Thursday, September 20, 12

Upload: tu-delft-transport-institute

Post on 20-Jan-2015

195 views

Category:

Technology


1 download

DESCRIPTION

Transport Thursday is an initiative of the TU Delft Transport Institute. Dr.ir. Hans van Lint gave a presentation on Innovations in Data Collection and Applications of Serious Gaming, all related to the field of transport. Enjoy!

TRANSCRIPT

Page 1: Transport Thursday 20 September 2012 - Why we need a virtual traffic and travel laboratory to research drive and travel behaviour

Challenge the future 1

Transport ThursdaysAn initiative of the TU Delft Transport Institute

Dr. ir. Hans van Lint (CEG/T&P)

Why we need a virtual Traffic and Travel Laboratory

Thursday, September 20, 12

Page 2: Transport Thursday 20 September 2012 - Why we need a virtual traffic and travel laboratory to research drive and travel behaviour

Challenge the future 2

Innovations in Data Collectionand Applications of Serious Gaming

Prof. dr. Serge Hoogendoorn & Dr. Hans van Lint

With contributions from:Winnie Daamen, Mignon van den Berg, Victor

Knoop, Raymond Hoogendoorn and the researchers from National Institute of

Informatics, Tokyo

Thursday, September 20, 12

Page 3: Transport Thursday 20 September 2012 - Why we need a virtual traffic and travel laboratory to research drive and travel behaviour

Traffic and Transport Models

•Traffic operations result from human decision making and complex multi-actor interactions at different behavioral levels )

•Human behavior is ‘not easy to capture and predict’

•System is highly complex, non-linear, has chaotic features, etc.

•The traffic flow and transportation theory does not exist: it is not Newtonian Physics or thermodynamics!

•Challenge is to develop theories and models that represent and predict operations sufficiently accurate for application at hand

•But how is this achieved? Induction vs deduction...

…for Distinction Sake, a Deceiving by Words, is commonly called a Lye, and

a Deceiving by Actions, Gestures, or Behavior, is called Simulation…

Robbert South (1643–1716)

Thursday, September 20, 12

Page 4: Transport Thursday 20 September 2012 - Why we need a virtual traffic and travel laboratory to research drive and travel behaviour

Challenge the future

Deduction

•Starts with an axiom, an assumed truth, a theory (which come from an observations, logic, other theories)

•Typical in (theoretical) physics, mathematics

•Example: special theory of relativity (Einstein postulated that the speed of light is the same for all observers, regardless of their motion relative to the light source – observations proved him right)

4

Theory / theories

An assumed truth

Confirmation / rejection

Observations / predictions

Hypothesis

On the basis of these theories / truths

Testing / analyzing

Qualitative (math) / quantitative (sim)

Modeling approaches

Thursday, September 20, 12

Page 5: Transport Thursday 20 September 2012 - Why we need a virtual traffic and travel laboratory to research drive and travel behaviour

Challenge the future

Induction

•Starts with observations (phenomena, patterns, etc.)

•Typical in social sciences and biology

•Example: Darwin’s theory of evolution by natural selection (Darwin observed populations finks diverging in different habitats and postulated natural selection as the motor – modern genetics, biology and many, many other scientific disciplines proved him right)

5

Observations

An assumed thruth

New theory

Until falsified...

Tentative hypotheses

About underlying relations / theories

Testing / operationalizing

Qualitatively / quantitatively

Modeling approaches

Thursday, September 20, 12

Page 6: Transport Thursday 20 September 2012 - Why we need a virtual traffic and travel laboratory to research drive and travel behaviour

Challenge the future

Traffic and Transportation Theory?

•Traffic flow theory is largely based on induction (generalizing fom empirical or experimental observations)

•Our theories and models are as good as the quality of their predictions (and should be assessed with that in mind!)• Do they predict the key phenomena

we observe in the real world?

• Do they incorporate a (mathematical) structure that provide insight into how these phenomena emerge?

•Let us recall some of these phenomena...

6

Inductive or deductive?

Observations

Discoveries Theories

Predictions

Thursday, September 20, 12

Page 7: Transport Thursday 20 September 2012 - Why we need a virtual traffic and travel laboratory to research drive and travel behaviour

Challenge the future

Bruce Greenshields...

•First traffic data collection using cameras and may hours of manual labour...

•Studied relation between average vehicle speeds and vehicle density (= average distance-1) and found an important relation

7

...and the discovery of the Fundamental Diagram

Thursday, September 20, 12

Page 8: Transport Thursday 20 September 2012 - Why we need a virtual traffic and travel laboratory to research drive and travel behaviour

Challenge the future

Bruce Greenshields

•Decreasing relation between speed and density

•When speed decreases, drivers drive closer together

8

...and the discovery of the Fundamental Diagram

Thursday, September 20, 12

Page 9: Transport Thursday 20 September 2012 - Why we need a virtual traffic and travel laboratory to research drive and travel behaviour

14.2 14.4 14.6 14.8 15 15.23000

3500

4000

4500

5000

q (v

tg/u

)

t (uur)

v (k

m/u

)

4120 veh/h

3600 veh/h

Another important phenomenon

•Capacity = max hourly rate of vehicles passing a cross-section (e.g. Coentunnel = 4120 veh/h)

•Before on-set of congestion, capacity is 10-15% higher than after on-set of congestion

•E.g. data A10 Coentunnel shows reduction to 3600 veh/h after congestion onset (-12.6%)

•Empirically established using data collected at cross-sections using inductive loops + smart analysis

Coentunnels101 location

Thursday, September 20, 12

Page 10: Transport Thursday 20 September 2012 - Why we need a virtual traffic and travel laboratory to research drive and travel behaviour

Pedestrian flow capacity drop

•Capacity drop also occurs in pedestrian flow

•Faster = slower effect

•Pedestrian experiments (TU Dresden, TU Delft) have revealed that outflow reduces substantially when evacuees try to exit room as quickly as possible (rushing)

•Capacity reduction is caused by friction and arc-formation in front of door due to increased pressure

•Capacity reduction causes severe increases in evacuation times

Thursday, September 20, 12

Page 11: Transport Thursday 20 September 2012 - Why we need a virtual traffic and travel laboratory to research drive and travel behaviour

Challenge the future 11

Traffic instabilities

•Field data analysis (top figure) and physical experiments (bottom movie) show that in certain density regimes, traffic is unstable

•Small disturbances amplify as they travel from one vehicle to the next

•Eventually, disturbance grows into so-called wide moving jam, moving upstream in opposite direction of traffic at speed of 18 km/h

•Outflow of wide-moving jam is about 30% less than free flow capacity

Thursday, September 20, 12

Page 12: Transport Thursday 20 September 2012 - Why we need a virtual traffic and travel laboratory to research drive and travel behaviour

Challenge the future 12

The story so far...

•Traffic and transportation theory is largely based on induction

•Network traffic flow is characterized by many emergent phenomena (capacity drop, instability, hysteresis, Braess, etc.)

•Theories and models need to reproduce and predict these, and may provide additional explanations on why and how these phenomena occur

•Examples show how the discovery of these phenomena often coincide with innovations in data collection and analysis

Thursday, September 20, 12

Page 13: Transport Thursday 20 September 2012 - Why we need a virtual traffic and travel laboratory to research drive and travel behaviour

Data Collection

Overview of TechniquesRecent innovations and resultsBenefits and drawbacks

Thursday, September 20, 12

Page 14: Transport Thursday 20 September 2012 - Why we need a virtual traffic and travel laboratory to research drive and travel behaviour

Challenge the future

Overview

•Roughly three classes of data collection:• Field observations (including revealed preference / choice)

• Controlled physical experiments

• Stated preference or choice / hypothetical situations

•Provide examples (including recent innovations) illustrating new insights stemming from analyzing these data

•Briefly discuss main benefits and drawbacks along four main dimensions: data validity, collection cost, quality and richness of collected data, and controllability of conditions

•Focus on vehicular traffic and pedestrian flows...

14

Data collection categories and some recent examples

Thursday, September 20, 12

Page 15: Transport Thursday 20 September 2012 - Why we need a virtual traffic and travel laboratory to research drive and travel behaviour

Field Observations

Monitor traffic and travel behavior as it occurs ‘naturally’

Thursday, September 20, 12

Page 16: Transport Thursday 20 September 2012 - Why we need a virtual traffic and travel laboratory to research drive and travel behaviour

Challenge the future

Field data collection

•Traditional approaches include: • Video, film, photo (Greenshields, Treiterer & Myers), which are expensive,

time-consuming, and yield small data sets• Cross-sectional data using single or double loops (inductive, pneumatic),

relatively cheap, big data sets, course data

•Technological innovations in data collection and processing has resulted in a revolution in field data collection (GPS, GSM, Bluetooth, radar, image processing, natural driving, etc.)

•Example innovation VIDI program “Tracing Congestion Dynamics”• Aim: insight into spatial-temporal behavior of all vehicles in flow aiming to

find explanations for observed phenomena (e.g. capacity drop)

• Collecting vehicle trajectory data from aerial platform • Application of model-based statistical data analyses and resulting insights

16

Monitor traffic and travel behavior as it occurs naturally

Thursday, September 20, 12

Page 17: Transport Thursday 20 September 2012 - Why we need a virtual traffic and travel laboratory to research drive and travel behaviour

Challenge the future 17

Aerial Vehicle Trajectory Data Collection

•Data collection system for collecting high-frequency images from the air (helicopter)

•Algorithms for stabilization of images and geo-referencing

•Vehicle detection and tracking, resulting in high-resolution data on revealed driving behavior (long + lat)

•15-30 min of data, 500 m roadway, 15 Hz, 40 cm resolution, all vehicles!

•Multiple data sets for variety of circumstances (congestion, merges, incidents, etc.)

Thursday, September 20, 12

Page 18: Transport Thursday 20 September 2012 - Why we need a virtual traffic and travel laboratory to research drive and travel behaviour

Challenge the future

Data analyses

•Analyses of vehicle trajectory datasets have yielded many new empirically underpinned insights in driving behavior:• Empirical evidence for piecewise constant acceleration (psycho-spacing models) profiles instead of continuous acceleration profiles (traditional car-following models)

• New perspectives on merging behavior: gap-seeking instead of gap-acceptance behavior (used in most models!)

• Empirical evidence of multi-anticipation (drivers have multiple leaders)

• Behavioral adaptation effects, e.g. near incidents (large changes in observed driving behavior)

18

and new insights...

Thursday, September 20, 12

Page 19: Transport Thursday 20 September 2012 - Why we need a virtual traffic and travel laboratory to research drive and travel behaviour

Challenge the future

Model-based behavioral analyses

•Generalization of Maximum-Likelihood estimation allowing estimation and statistical analyses of model parameters for individual drivers

•Example (Intelligent Driving Model):

•Parameters identified using G-ML approach showing:• Large heterogeneity in driving population in terms of parameter values

• Model represents adequately revealed behavior part of population

19

Statistical analyses by means op model-based approaches

ddt

vi(t) = a ⋅ 1−

vi(t)

v0

⎝⎜

⎠⎟

4

−s *(v

i,Δv

i)

si

⎝⎜

⎠⎟

2⎛

⎝⎜⎜

⎠⎟⎟

s *(v

i,Δv

i) = s

0+Tv

i+v

iΔv

i/2 a ⋅b

i

v

i(t)

v

i−1(t)

Δv

i(t)

s

i(t)

i −1

Thursday, September 20, 12

Page 20: Transport Thursday 20 September 2012 - Why we need a virtual traffic and travel laboratory to research drive and travel behaviour

Challenge the future

Variability in driving behavior

20

•Estimation provides insight into distribution of parameters across the driving populations

•Large variations can be found (driver heterogeneity)

IDM Estimation Results

Thursday, September 20, 12

Page 21: Transport Thursday 20 September 2012 - Why we need a virtual traffic and travel laboratory to research drive and travel behaviour

Challenge the future

All models are wrong!

•Method has been applied to compare different types of models

•Best fitting model is different for each driver (picture shows how many times model performed better than all others)

21

But some are more wrong than others

0

13

25

38

50

BexeliusHelly

LenzGipps

IDM

Thursday, September 20, 12

Page 22: Transport Thursday 20 September 2012 - Why we need a virtual traffic and travel laboratory to research drive and travel behaviour

Challenge the future 22

Next steps in aerial data collection•Data collection via helicopter is very expensive

•Use of UAS (Unmanned Airborne Systems) to collect data from the air

Thursday, September 20, 12

Page 23: Transport Thursday 20 September 2012 - Why we need a virtual traffic and travel laboratory to research drive and travel behaviour

Challenge the future

Pro’s and cons

Advantages•Representative data (for the situation encountered)

•Realistic conditions and behavioral responses

Disadvantages•Cost and effort to collect and process data (depending on data collection technique)

•Uncertain if relevant situation occurs and (controllability of conditions)

•Limited info about subjects

•Intervening not always possible

23

Empirical research by mean of field observations

Thursday, September 20, 12

Page 24: Transport Thursday 20 September 2012 - Why we need a virtual traffic and travel laboratory to research drive and travel behaviour

Physical Experiments

Conducting Experiments in a real or in a representation of physical environment

Thursday, September 20, 12

Page 25: Transport Thursday 20 September 2012 - Why we need a virtual traffic and travel laboratory to research drive and travel behaviour

Challenge the future

Controlled condition experiments

•Use real or representation of physical environment •Subjects are instructed (or otherwise made) to perform certain tasks within this environment (experiments)

•Tasks execution is closely monitored•Different examples using either the real objects of study:• Driving experiments• Pedestrian experiments• Large evacuation experiment at

Euroborg to test lane reversal

...or using ‘proxies’ for the objects:• Rice, grain, ants, slaters, mice, etc.

25

Courtesy of Majid Sarvi

Physical experiments in real environments

Thursday, September 20, 12

Page 26: Transport Thursday 20 September 2012 - Why we need a virtual traffic and travel laboratory to research drive and travel behaviour

Challenge the future

Pedestrian experiments

•Controlled experiments by TU Delft (2002) and Hermes (2009)

•Insight into dynamic lane formation process

26

What happens if two groups of pedestrians meet head-on?

Thursday, September 20, 12

Page 27: Transport Thursday 20 September 2012 - Why we need a virtual traffic and travel laboratory to research drive and travel behaviour

Challenge the Future

Pro’s and cons

Advantages•Controllability of conditions

•Possibility to intervene, test new designs, etc.

•Possibility to collect information about subjects (at least regarding overall population)

•Data collection and processing relatively straightforward (e.g. using hats makes tracking of pedestrians from video easy)

Disadvantages•Effort & cost can be substantial

•Level of realism not so high as for empirical observation (?)

•Safety of subjects, ethics

27

Thursday, September 20, 12

Page 28: Transport Thursday 20 September 2012 - Why we need a virtual traffic and travel laboratory to research drive and travel behaviour

Stated Preferences

Observing choice behavior in Hypothetical Scenarios

Thursday, September 20, 12

Page 29: Transport Thursday 20 September 2012 - Why we need a virtual traffic and travel laboratory to research drive and travel behaviour

Challenge the future

Stated-preference or choice

•Present (hypothetical) choice situations to subjects and record their decision behavior for a variety of circumstances

•By considering different situations, attributes of choice behavior can be systematically investigated

•Approach has been used for years and years: many, many examples can be given!

•Examples generally pertain to tactical / strategic behavior studies (travel behavior decisions)

•Recent innovations use web-based techniques and semi-interactive environments to study travel behavior

29

Hypothetical choice scenarios

Thursday, September 20, 12

Page 30: Transport Thursday 20 September 2012 - Why we need a virtual traffic and travel laboratory to research drive and travel behaviour

Challenge the future

Web-based choice experiment

•Dynamic evacuation scenario, choice to stay / leave, mode and route choice

30

Example evacuation behavior study

Thursday, September 20, 12

Page 31: Transport Thursday 20 September 2012 - Why we need a virtual traffic and travel laboratory to research drive and travel behaviour

Challenge the future

Example results

•361 participants took part in experiment, 3 dynamic scenarios per respondent

•Analysis of choice behavior undera large variety of conditions, personal characteristics, etc. (example resp. curve male / female)

•Sequential binary choice model (stay / leave at this period), separate for car and bus modes

•Estimated using Discrete Choice models, showing plausible results, but data validity still is an issue

31

Departure time profile

Thursday, September 20, 12

Page 32: Transport Thursday 20 September 2012 - Why we need a virtual traffic and travel laboratory to research drive and travel behaviour

Challenge the future

Pro’s and cons

Advantages•Controllability of conditions

•Possibility to test interventions

•Possibility to collect detailed information about subjects

•Data collection and processing straightforward

Disadvantages•Level of realism: stated vs revealed behavior

•No pay-off or loss, choices have limited consequences

32

Stated preference research

Thursday, September 20, 12

Page 33: Transport Thursday 20 September 2012 - Why we need a virtual traffic and travel laboratory to research drive and travel behaviour

Challenge the future

Comparing the methods...

33

0"1"2"3"4"5"6"7"8"9"10"Validity"

Controllability"

Data"quality"/"richness"

Costs"

Field"observaBons"

Experiments"

Stated"Preference"

...with some indicative numbers

Thursday, September 20, 12

Page 34: Transport Thursday 20 September 2012 - Why we need a virtual traffic and travel laboratory to research drive and travel behaviour

Data collection for Behavioral Modeling - ICEM 2012 34

EverscapeBehavioral data collection by means of 3D Internet Technology

Evacuation of a virtual island...

Thursday, September 20, 12

Page 35: Transport Thursday 20 September 2012 - Why we need a virtual traffic and travel laboratory to research drive and travel behaviour

Challenge the future

Everscape

•Combine advantages of different data collection methods (field data / RP, SP, controlled exp.)

•Provide an ‘as realistic experience’ as possible using 3D environment

•Multi-user platform to study emergent behavior, circumvent use of bots for now

35

•Joint consideration of traffic and travel behavior (driver / travel simulator)

What is it?

Thursday, September 20, 12

Page 36: Transport Thursday 20 September 2012 - Why we need a virtual traffic and travel laboratory to research drive and travel behaviour

Challenge the future

Virtual environments

•Virtual environments have been proposed as alternative means to collect data for emergency conditions

•Two examples:• Virtual Reality tool BART (M. Kobes, see ICEM 2009)

• Driving simulator experiments

•Validation studies show (relative) validity of environment

36

Thursday, September 20, 12

Page 37: Transport Thursday 20 September 2012 - Why we need a virtual traffic and travel laboratory to research drive and travel behaviour

Driver Simulator Studies

•Different studies have concluded on relative validity of driving simulators

•Use of driving simulator to study behavioral adaptation to rare conditions (incident, bad weather, stress)

•Detailed data on participants (physiology), including heart rate, facial expression, respiration

•Cross-comparing data with field observations (incidents) reveals strong similarities between data sources

•Key issue remains unrealistic behavior of the bots...

Thursday, September 20, 12

Page 38: Transport Thursday 20 September 2012 - Why we need a virtual traffic and travel laboratory to research drive and travel behaviour

Challenge the future

•Unity engine for realistic video and sound experience (important for providing sense of danger stemming from earthquake / tsunami)

•Complete freedom of movement and free choice behavior

•User interaction via visual cues, communication between users via chat function (global)

•HMI via keyboard and mouse

•Detailed data collection on activities of participants•Internet based data collection (no need to sit in one room)

•Free version up to 100 participants; commercial version > 1000!

Properties of Everscape Virtual Lab

38

Some technical aspects...

Thursday, September 20, 12

Page 39: Transport Thursday 20 September 2012 - Why we need a virtual traffic and travel laboratory to research drive and travel behaviour

Challenge the future

•Pilot study objective: • to investigate potential of 3D Internet Technology to collect Behavioral Data, focussing on user-experience and tool validity

•Two pilot studies:• 2 experiments @ TRB (January 24, 2012) at George Washington University courtesy of Prof. Samer Hamdar

• 1 experiment @ PLATOS (March 14, 2012)

Pilot studies

39

Thursday, September 20, 12

Page 40: Transport Thursday 20 September 2012 - Why we need a virtual traffic and travel laboratory to research drive and travel behaviour

Overview of Island and experimental set-up

Thursday, September 20, 12

Page 41: Transport Thursday 20 September 2012 - Why we need a virtual traffic and travel laboratory to research drive and travel behaviour

train short route

long route

bridge

tsunami

heliport

concert

Overview of Island and experimental set-up

Participants arrive by helicopter

Participants drive to concert by car

During concert, earthquake occurs

followed by newsflash

People are free to evacuate whenever

they want...

Thursday, September 20, 12

Page 42: Transport Thursday 20 September 2012 - Why we need a virtual traffic and travel laboratory to research drive and travel behaviour

Impression of TRB Experiment

Thursday, September 20, 12

Page 43: Transport Thursday 20 September 2012 - Why we need a virtual traffic and travel laboratory to research drive and travel behaviour

Impression of experience from TRB experiment

•Video from experiment in which 26 people participated, conduction during TRB Annual Meeting 2012

Thursday, September 20, 12

Page 44: Transport Thursday 20 September 2012 - Why we need a virtual traffic and travel laboratory to research drive and travel behaviour

Challenge the future

Data collected

•Static / personal data (gender, age, education, gaming experience, etc.)

•Dynamic data (position, facing direction, speed, braking / acceleration, mode) for each second

•Visual field information (what did people see?)

•Motivation and user-experience (from survey / discussions after experiment)

Heli-platform

Concert

44

Using Everscape

Thursday, September 20, 12

Page 45: Transport Thursday 20 September 2012 - Why we need a virtual traffic and travel laboratory to research drive and travel behaviour

Challenge the future

Herding?

•Emergent behavior stemming from interaction?

•About 32% of the participants said that their behavior was influenced by others: • ‘Because we decided to take the train together’

• ‘I just followed someone’

• ‘I wanted to be in a car before the others so I was sure that I was the first one on the road’

•One person missed the earthquake and news item BUT saw everyone suddenly run away and followed!

45

Emergent user-behavior analyses

Thursday, September 20, 12

Page 46: Transport Thursday 20 September 2012 - Why we need a virtual traffic and travel laboratory to research drive and travel behaviour

Challenge the future

•Preliminary analyses of departure time choice

•Estimation of sequential binary choice model to study incentives for people to start their evacuation

•Participants decide to leave in period t based on utility of leaving vs. utility of staying:

•Few data-points, so expectations regarding statistical significance are low...

Modeling example

U

i(t) = ASC + β

1⋅n

leaving(t)+ β

2⋅news(t)+ β

3⋅earthquake(t)

46

Emergent user-behavior analyses

t

t+1

t+2

:

:

evacuate stay

:

:

( )p t

( 1)p t +

1 ( )p t−

1 ( 1)p t− +

( 2)p t + 1 ( 2)p t− +

evacuate stay

Thursday, September 20, 12

Page 47: Transport Thursday 20 September 2012 - Why we need a virtual traffic and travel laboratory to research drive and travel behaviour

Challenge the future

Modeling?

•Impact of earthquake could not be established

•Preliminary conclusions from model estimation: • People have tendency to stay

• Newsflash has impact of decision to leave

• Large impact of number of people leaving on decision to leave

mean t-value

ASC 0.799 1.94

BETA1 -0.426 -2.42

BETA 2 -0.688 -1.76

47

Emergent user-behavior analyses

Thursday, September 20, 12

Page 48: Transport Thursday 20 September 2012 - Why we need a virtual traffic and travel laboratory to research drive and travel behaviour

Challenge the future

Level of realism

What’s good:•Realistic experience, sense of haste / pressure

•Earth quake is realistic

•Awareness of others and their behavior

•Controlling avatar satisfactory

What to improve:•Sound: news item, screaming

•Controlling car

•Different experience for those with gaming exp.

•Oral chat, and chat radius

•Ability to travel together (e.g. get in car together)

48

Thursday, September 20, 12

Page 49: Transport Thursday 20 September 2012 - Why we need a virtual traffic and travel laboratory to research drive and travel behaviour

Challenge the future

Towards a VTT Lab

•Large scale experiments, in depth consideration of experimental design aspects, etc.

•Improvements in HMI (depending on aspects under investigation), e.g. by interfacing with driving simulator

•Inclusion of bots with realistic driving / travel behavior (self-learning): fusion of simulator with advanced simulation

•Inclusion of data lab (fusion of other types of data, such as empirical data)

•Extensive tool validation!!!

49

Proposed improvements / innovations:

Thursday, September 20, 12

Page 50: Transport Thursday 20 September 2012 - Why we need a virtual traffic and travel laboratory to research drive and travel behaviour

Learning from gaming

•Collaboration with gaming research groups (GATE)

•Modeling crowd behavior (tactical, operational) to provide for convincing behavior

•Do these models need to be predictively valid?

•Computational efficiency is important issue, both for walking behavior and for route choice behavior (density based routing)

Thursday, September 20, 12

Page 51: Transport Thursday 20 September 2012 - Why we need a virtual traffic and travel laboratory to research drive and travel behaviour

Closure

Challenges in Data Collection

Thursday, September 20, 12

Page 52: Transport Thursday 20 September 2012 - Why we need a virtual traffic and travel laboratory to research drive and travel behaviour

Challenge the futureChallenge the future

In conclusion

•Advances in data collection and analyses techniques have lead to important contributions in traffic and transportation theory•Discovery of new phenomena driving network dynamics•Insight into underlying mechanisms of these phenomena

•Capture these characteristics in models that are predictively valid and help us understand mechanisms better

•Contemporary models can (reasonably) describe phenomena such as capacity drop, traffic instabilities, lane formation, but predictive validity is still an issue (unsatisfactory accuracy of model in situations in which model is not calibrated)

•Innovations in data collection / analysis required to further our understanding of underlying behavioral processes and adaptions: we just scratched the surface!

Lessons learnt

52

Thursday, September 20, 12

Page 53: Transport Thursday 20 September 2012 - Why we need a virtual traffic and travel laboratory to research drive and travel behaviour

Challenge the future

Scientific challenges

•Development of new data collection, storage & retrieval, and processing platforms and tools

•Furthering data analyses techniques, e.g. to combine data stemming from difference sources (e.g. field data and virtual data; similar to combining RP and SP data)

•Development of VTT Lab:• Immersive experience leading to useful data?

• Convincing (adaptive!) behavior of bots for virtual environments (driving simulator) to improve validity of data collection tool, while maintaining computational efficiency

• What is ‘convincing’? Is real behavior convincing?

• Extensive user surveys using methods from gaming science

53

Towards a VTT Lab (Virtual Traffic and Travel Laboratory)

Thursday, September 20, 12