Session 22 Trent Victor

36
Trent Victor SAFER and Volvo Technology 2009-01-08 - some preliminary project results
  • date post

    12-Sep-2014
  • Category

    Technology

  • view

    342
  • download

    0

description

 

Transcript of Session 22 Trent Victor

Page 1: Session 22 Trent Victor

Trent Victor

SAFER and Volvo Technology

2009-01-08

- some preliminary project results

Page 2: Session 22 Trent Victor

Background

Page 3: Session 22 Trent Victor

Background Motivations• What causes accidents?

– Greatly increased knowledge of driver behavior, ”the Human Factor”, as a contributor to crashes.

– Study how driver interacts with vehicle, safety systems, road, traffic, weather, etc

• What can we do about them? – Evaluation of new technology (e.g. active safety systems)– Development of new technology and countermeasures based on the

findings– How to improve crash-avoidance behaviors

“Naturalistic driving studies are defined as those undertaken using unobtrusive observation or with observation taking place in a natural setting” (Dingus, et al. 2006).

Field Operational Tests are defined as “a study undertaken to evaluate a function, or functions, under normal operating conditions in environments typically encountered by the host vehicle(s) using quasi-experimental methods” (FESTA, 2008)

Naturalistic Field Operational Tests combine both – this is the focus of SAFERs partners

Page 4: Session 22 Trent Victor

“Naturalistic driving (ND) data collection is used to assess the relationship of (permanent and transient) Driver-Vehicle-Environment (DVE) factors with crash risk, driving behavior, and countermeasure effectiveness.” Naturalistic Field Operational Tests

DVE

Fac

tors

Driver Factors- Permanent: Age, Experience, Style…- Transient: Drunk, Tired, Distracted…

Vehicle Factors- Permanent: Vehicle type, Spec…-Transient: Active Safety, Nomadic Dev…

Environment Factors- Permanent: Speeds, Road type…- Transient: Weather, Lighting…

Naturalistic driving (ND) data collection- Natural driving, no special instructions, own vehicles, no experimenter present, unobtrusive data collection instrumentation…

is used to

assess the relationship

between

Crash Risk- Relative risk, Population attributablerisk…

Driver Behavior- Control behavior (lat, long), Attention, Decisions, Usage/adoption, Event involvement…

Countermeasure effectiveness- Active Safety, Road treatments, etc

Page 5: Session 22 Trent Victor

UMTRI: FCWUMTRI: LDW

Japan1: Crossing RoadJapan2: Frontal Collision

Japan3: DrowsinessJapan4: Near missVTTI: 100-Car study

SAFER: TSS FOT

Page 6: Session 22 Trent Victor

0

20

-20t

Eyes on road

Eyes off road

Steering angle

Lane exceedence Lane position

Lamp pole

Warnhere?

Warnhere?

Acthere?

Informhere?

Page 7: Session 22 Trent Victor

Time to Lateral Collision

Tim

e to

Lon

gitu

dina

l Col

lisio

n

OO

O

O

Infinity

Infin

ity

X

Collision Left Collision Right

Col

lisio

nFr

ont

Col

lisio

nB

ack

CrashNear CrashIncident

Events of Interest

Undisturbed Passages

Directly Safety Related:

Indirectly Safety Related:

Crash avoidancebehaviors

Page 8: Session 22 Trent Victor

Fatal

Injury(light/moderate/severe)

Police-ReportedProperty Damage Only

Non-police-reportedProperty Damage Only

Non-police-reportedPhysical contact or tire strike

Events of interest

Exposure/occurancy

Near crash

Incidents

Safe

tysy

stem

X

Event of relevance for research (e.g. Accidentology)Event of relevance for evaluation of Safey System X Crashes as defined in databases

(police/ambulance-reported)15Crashes as defined in the

100-Car study

5 x police-reported (PR) Crashes

67

50 x PR crashes761

550 x PR crashes8295

FOT/NDS

Page 9: Session 22 Trent Victor

SHRP2 (USA)• Extensive observations of driving

behavior• >2500 cars for 2 yrs• Active Safety subset of 500 cars…

• Representative sample of crash data and near-crash data

• Databases available for “the next generation of traffic safety researchers”

Track 1: In-Vehicle StudyS06: Technical Coordination and Independent Quality Assurance for

Field Study--$3M

2008 20112009 2010

S05: Design of the In-Vehicle Driving Behavior and Crash Risk Study--$3M

2007

S07: In-Vehicle Driving Behavior Field Study--$28M

Study Design & Field Data Collection

2012

Roadway Data

Analysis

Track 2: Site-Based Study

S08: Analysis of In-Vehicle Field Study Data and Countermeasure Implications--$4M

(multiple awards, different letting schedules)

S09: Site-Based Video System Design and Development--$1M

S01: Development of Analysis Methods Using Recent Data--$1.5M (multiple

awards, two phases)

S03: Roadway Measurement System

Evaluation $0.5M

S02: Integrate Methods and Develop Analysis Plan--$0.5M

Revised December 2007

S11: Analysis of Site-Based Field Study Data and Countermeasure Implications--$2M

(contingent project)

S10: Design and Conduct of the Site-Based Field Study--$11M (contingent project)

S04: Acquisition of Roadway Information--$3.5M

Study Design & Field Data Collection

Page 10: Session 22 Trent Victor

Consumer SystemsJapanese systems

(insurance-driven - taxi, fleet)

Pay-as-you-drive(insurance-driven)

Risk management systems(for e.g. fleets, parents)

Remote diagnosticsand fleet management

Page 11: Session 22 Trent Victor

Field Operational Test (FOT) start-up at SAFER

2006 (3MSEK) 2008 (100MSEK)

Sweden

USA

SAFER

Europe

TSSFOT (2*)Methodology

FESTAMethodology

EuroFOT (150*)Impacts

FOTNetCoordination

SeMiFOT (18*)Methodology

INTENDMethodology

EstablishFOT

BASFOTCompetence build-up

2007

UMTRI UMTRI SHRP2

* Swedish vehiclesProjectCompetence Proposal

Page 12: Session 22 Trent Victor
Page 13: Session 22 Trent Victor

• Part of Sweden-Michigan partnership agreement

• Main Goals: – to further develop the Naturalistic FOT method into a powerful

tool fora) Accidentologyb) Evaluation of safety, efficiency, and usage &

acceptancec) Countermeasure innovation and development

• 18 vehicles in Sweden and 2 vehicles in USA, ca 6 monthsdata collection, duration Jan 2008-June 2009

Page 14: Session 22 Trent Victor

WP2 – Methodology and FOT Design

Page 15: Session 22 Trent Victor

WP2 – Methodology and FOT Design• Identification of function and vehicles

– The selection of systems is more guided by what systems are available and what systems the manufacturers wanted to include in this project

– On-market vehicle-integrated systems and one after market system• Definition of objectives, hypotheses, and performance indicators for each function

– Next slides• Specification of experimental procedures

– FESTA Handbook– Study plan was submitted for ethical review – Data and personal integrity, data ownership

and sharing, much more complicated and multi-faceted than assumed. Many legal issues, e.g. responsible for filming.

Decision from ethical committee in Gothenburg – this study does not need ethical approval.

– Flexibility in experimental procedures is rather constrained by practical issues, OEM, and safety requirements

– Vehicles and drivers selected from manufacturers or manufacturer-associated companies –Primary car drivers (and family members) vs truck drivers

– Comparable scenarios in the baseline data, when the function is turned off, and in the treatment data, when the function is turned on. Changes over time.

– AB design, no baseline for some functions (e.g. ESC)– a relatively large number of questionnaires

Page 16: Session 22 Trent Victor

Accide

ntolog

y

ACCLD

WBLIS FCW

ESCIW

Saf

ety

Atte

ntio

nU

sage

Acc

epta

nce

Crash-Relevent Event Analysis-Multiple regression etc, relatingPrecursor, Outcome, Mediating factors

Visual Behavior Analysis-Glance behavior ”function”, Distractionevents

Usage Analysis-Quantify usage in select situations

Acceptance Analysis-Quantify acceptance, relate to usage

25 50 65 105115

19 59 19

11 41 26 6

45 75 35 75

CR-Events-Prevented Analysis-””What-if” no system acted?” analysis

30

4

0

45

30

4 4

0

315

290

109

66

160 130 180 45 135 86 44(21%) (17%) (23%) (6%) (17%) (11%) (6%)

(40%)

(37%)

(14%)

(8%)

780(100%)

Page 17: Session 22 Trent Victor

Hypothesis example (ACC)

Page 18: Session 22 Trent Victor

Conclusions on Hypothesis Prioritizations

• Safety and Attention analyses should be prioritized as they received 77% of the prioritization points, whereas the Usage and Acceptance analyses received 22%.

• LDW, Accidentology, ACC, and FCW should be the prioritized applications of the analysis.

• Further prioritizations:– Within Safety analyses, prioritize analysis of crash-relevant events (i.e.

kinematic- and system triggers)

– Within Attention analyses, prioritize analysis of eyetracker data in selected situations

– Within Usage analyses, prioritize analysis of usage for the LDW, ACC, and ESC functions.

– Within Acceptance analyses, prioritize analysis of acceptance questionnaires.

Page 19: Session 22 Trent Victor

WP3 – Data Management

Page 20: Session 22 Trent Victor

Virginia Tech Naturalistic Driving Equipment (SHRPII study)

Presentation by Tom Dingus as SHRP 2 Safety Research Symposium, July 17-18,2008, Washington, DC

2600 vehicles!!

Page 21: Session 22 Trent Victor

Virginia Tech Naturalistic FOT Equipment (100-car study)

Page 22: Session 22 Trent Victor

UMTRI Naturalistic FOT Equipment

Page 23: Session 22 Trent Victor

PC

SAFER Data Acquisition SystemExtra ”external” sensorsAccelerometersEyetrackers – SeeingMachines/SmartEye (13units total)Lanetracker/ForwardDistVel – MobilEye (15 units)

GPS (1 Hz)

CANSteering Wheel AngleTurn IndicatorGear Level PositionAccelerationsEtc …..

Video (Analogue)6 Cameras in total

Hard drives

Two different data acqustion

systems evaluated

Page 25: Session 22 Trent Victor

Database and Storage

• Very large data volumes!• SeMiFOT:

– Video: 8 Terabyte– Data: 1 Terabyte

• euroFOT (Sweden only):– Video: 50-100 Terabyte– Data: 6 Terabyte

Video

Data

Page 26: Session 22 Trent Victor

Analysis tools

Video

Data– Direct database use/searching– Event identification– Synchronized data with video– Easy manual and automatic

annotations

[Show video]

Page 27: Session 22 Trent Victor

WP4 – Vehicle and Test Management

Page 28: Session 22 Trent Victor

Current status:• 3 VTEC – 2 trucks running• 3 SAAB – 2 cars running• 4 SCANIA – currently installing • 8 VCC-6-7 cars running with TSS-FOT

logger

Some aspects:• Installation and verification• Pickup of data in vehicles• On-line quality control• Hotline and support organization• Data uploading

Page 29: Session 22 Trent Victor

WP5 – Evaluation of Methodology

Page 30: Session 22 Trent Victor

WP5 – Evaluation of Methodology

• Consultations with UMTRI, SHRP2, Guestresearcher visit from IOWA (SHRP2 Analysis), FESTA, FOT-NET, EuroFOT

• Daunting, complex task but there are some trueopportunities, e.g. eyetracker data, events-prevented analyses, etc.

Page 31: Session 22 Trent Victor

Collaboration with SHRP2

SHRP2 is within the Transportation Research Board (TRB) of the National Academy of Sciences (NAS)

1. SeMiFOT as a collaboration probect with SHRP2 – Loan staff visit to SHRP2, Technical Expert Group participation

2. Memorandum of Understanding regarding information exchange between NAS (TRB) and Sweden (SRA and VINNOVA for SAFER)

Return visit by SHRP2 to Sweden in Feb/March

Page 32: Session 22 Trent Victor

Conclusions

• Ongoing project, new methods and technology are being developed for the first time in Europe. Has more of a methods development character.

• Has given Sweden and SAFER partners a leading position in EU and internationally

• Good collaboration with Michigan (UMTRI)• Complex project in many regards

Page 33: Session 22 Trent Victor

Borderless research to save lives

www.chalmers.se/safer

[email protected]

Page 34: Session 22 Trent Victor

Experimental Design Phase

NaturalisticDriving Studies

(NDS)

NaturalisticFOT

OtherFOTs

Tools

AnalysisPhase

Method Chain in Relation to NDS & FOT

Other FOTse.g. test routes

Naturalistic FOT

NaturalisticDriving Studies

(NDS)

NDS

FOTs

Collection Phase

Leve

l of e

xper

imen

tal c

ontr

olLo

wH

igh

Diff

eren

t Ana

lysi

sG

oals

Page 35: Session 22 Trent Victor

Naturalistic Methodologyin Relation to Existing Methods

Naturalistic Methodology – Objective longitudinal data (high km), large number of cases, unobtrusive instrumentation, no

experimenter present, driving their own vehicles, tens to thousands of vehicles, etc

Experimental Field Studies – low km, short time-scale, Experimental control, specific routes, few cases, ca. (1-10 cars)

etc

Experimental Lab-, Simulator-, and Test Track Studies

In-depth studies of Pre-Crash behaviour, initiated by CrashEvents (e.g. on-site investigations and interviews)

Aggregated data of Pre-Crash behaviour, initiated by CrashEvents (e.g. questionnaires)

Enabled by new data collectiontechnology

Page 36: Session 22 Trent Victor

Factors Influencing Choice of Objectives

1. Opportunities• Study new issues, develop innovative methods,

2. Resources• Time (hrs and calendar)• People with the right competence

3. Diverging partner interests• Especially OEM constraints (e.g. y-data)

4. Data reduction limitations • Ease of implementation limited by technology,

difficulty of Performance Indicator calculation etc, manual data-reduction