SeMiFOT

SeMiFOT
Used in FOTs:	SeMiFOT
Perspectives of the FOTs:	Driver behaviour
Countries:	Sweden
Data
Distance travelled	250,000 km <br />
Hours of logging	5,000 hours <br />300,000 minutes <br />208.333 days <br />
Number of journeys
Number of events
Detections	14 vehicles / travellers
Indicators
In the raw data	Position, Speed, Headway, Vehicle control related, Fuel consumption, HMI related, Video
Calculated / derived	Speed related, Position related, Headway related
Enriched information	Speed limit
Test subjects
Professional drivers	Included
Test subjects	20
Equipped road sections / cross-sections
Vehicles	14
Vehicle types	Passenger car, Truck
Test setup
Start of field tests	20081101
Length of field tests	9 months <br />0.75 years <br />
Length of baseline phase
Road type	Urban streets, Urban main roads, Rural roads, Main roads and corridors
Weather and other conditions included
Type of experiment	Naturalistic driving
Design of experiment

Data sets acquired

	Can be shared as
Data set	Raw data	Aggregated data
HMI
Questionnaires
Map data
GPS
CAN
Radar
Driver video data

Data formats and possible standards

Data stored in mat format

Frequency of logging

10 Hz

Additional notes

SeMiFOT
General information
Type: Field operational test
Tested system/service: Autonomous Systems
Countries: Sweden	? test users
13 partners	14 vehicles
Active from 01/2008 to 12/2009
Contact
Project Reports
Trent Victor
trent.victor@chalmers.se
Chalmers University of Technology Sweden
Catalogue entries
Data catalogue	Tools catalogue
Data sets used in this FOT: SeMiFOT	The following tools were used in this FOT: No tool is linked to this FOT Click here to create a new tool or here to browse existing tools to link them to this FOT

SeMiFOT will further develop the Naturalistic Field Operational Test (FOT) method into a powerful tool in traffic safety research.

The naturalistic method involves collecting data continuously from a suite of vehicle sensors in order to assess safety in the interactions of driver, vehicle and environment. Environment sensing and video are essential for identifying near-collisions and other incidents, and for validating that intelligent vehicle systems (e.g. collision warning, lane departure warning and intelligent speed adaptation) perform as expected.

SeMiFOT will focus on the tools in the methodology chain (data acquisition-data storage-data analysis) needed to perform a Naturalistic FOT. These tools will be evaluated on a number of selected in-vehicle and cooperative systems. The requirements for large scale FOT’s will be analysed.

Details of Field Operational Test

Start date and duration of FOT execution

Start: Nov-Dec 2008

Duration: ca 6 months data collection

Link with other related Field Operational Tests

SeMiFOT was originally intended to be a pilot project for a larger Field Operational Test. In the end, several large scale FOTs were started (EuroFOT and TeleFOT). As a matter of interest for the reader, projects that were somehow inspired, built upon, or connected with the work done in SeMiFOT include FESTA, EuroFOT, FOT-NET, BasFOT (SAFER project), TeleFOT, DREAMi (SAFER-Japan project), and a continuation of SeMiFOT called SeMiFOT2.

Objectives

The SeMiFOT project aimed at implementing and developing the Naturalistic Field Operational Test (N-FOT) method as a method to understand crash causation and the effect of new safety systems. The intention was to test the N-FOT methodology at an intermediate level, and therefore it was not the intention to perform a full evaluation of the safety impact of the safety systems. SeMiFOT focused on a selection of issues within the “methodology chain”, from data acquisition to data storage to data analysis. SeMiFOT was envisioned as a sort of pilot test, or a methods development pre-cursor to full-evaluation projects such as the EuroFOT project. Methodology development was at the center of attention.

The main general goal for SeMiFOT was to further develop the Naturalistic FOT method into a powerful tool for (a) accident research, (b) evaluation of safety, efficiency, and usage & acceptance, and (c) countermeasure innovation and development.

Results

The main advances by SeMiFOT address challenges related to:

1. Technology and implementation

• Review and analysis of a wide range of commercially available data acquisition systems from USA, EU and Japan, and their potential use in collaboration with insurance companies
• In-house development of state-of-the-art data acquisition systems, data/storage, and analysis tools together with UMTRI
• Implementation of data handling and sharing procedures for a joint database sharing both open and proprietary data from 4 OEMs
• SeMiFOT also provided the opportunity for learing and development on a large variety of topics which were very practical or mundane. The time spent on data acquisition and data preparation for analyses far exceeded expectations and future gains in these areas are particularly useful

2. Analytical approach

• Development of the Crash-Relevant Events (CRE) analysis method, involving a new trigger approach, coding scheme, integration with accident statistics geographically, and a new CRE-based safety impact assessment methodology
• Development of a novel Events-Prevented analysis method performing "what-if" analyses emerging out of parameters from specific situations
• Development and implementation of visual behavior analyses of data from 13 eyetrackers, which revealed large quality difficulties in sensing and classification of eye/head movements in naturalistic data
• Implementation of map-matching and use of map-data in analysis of automatic speed camera
• Integration of the naturalistic FOT method into company development processes

Lessons learned

General

1. Data processing and measure calculation is where the majority of analysis time is spent
2. Preparing the data as soon as it has been collected, i.e. before uploading in the database
3. Should have continued the study design Work Package until the end of the project, to monitor experimental design & scientific quality issues, and to monitor data quality iteratively

NDS style analyses

1. Showed that it is no easy task to analyse naturalistic data
2. More work is needed to be able to efficiently analyse future datasets such as the SHRP2 and EuroFOT datasets
3. Valuable step in the development of the competence and methodology needed
4. Helped specify content for SeMiFOT2

FOT-style analyses

1. Successfull in:

The development of innovative methods (events-prevented, and CRE of Safety Impact)

The first application of automatic visual behavior analysis in an FOT

A successfull usage analysis which included the development of a statistical model

An evaluation of the acceptance of all systems, and of subjective methods (questionnaires and instruments)

2. Most of the analyses have encountered the problem of determining the baseline or what to use as a ground truth

3. Shown the importance of careful quality check of data - often a need to check the data manually

4. Analysis time has in many cases exceeded by far the expectations

Data Management

1. Development, testing and the decision process of data acquisition must be allowed time and it will cost more than anticipated
2. The analysis and verification process of the data that is collected and pre-processed should be prioritized and performed before vehicles are sent out.

Questionnaire design

1. Some instruments need update and modernisation (e.g. Sensation Seeking Scale)
2. The standardised questionnaire have been developed for addressing private drivers, not professionals
3. Few questionnaires have been validated in another language than English
4. The web-based solution does not appear to motivate the participants to elaborate their answers to open-ended questions to any higher degree than the traditional paper and pen solutions

Main events

2010.06.01 SeMiFOT Final Seminar

Financing

Summary, type of funding and budget

Overall

2 Mio EUR

Public

Partly financed by the Michigan Department of Transportation (MDOT) and by the Swedish Governmental Agency for Innovation Systems (VINNOVA)

Cooperation partners and contact persons

• Public Authorities: Swedish Road Administration

• Industry:
  • Vehicle Manufacturer: AB Volvo, Saab Automobile, Scania, Volvo Car Corporation
  • Supplier: Autoliv, Volvia

• Users:
• Universities: Chalmers University Technology
• Research Institutes: Swedish National Road and Transport Research Institute (VTI), Technical Research Institute of Sweden (SP), UMTRI

• Others (specify): Länsförsäkringar (Insurance group), Test Site Sweden / Lindholmen Science Park

Main Contact persons

   Trent Victor
   SAFER
   trent.victor@chalmers.se
   +46 31 322 66 51

   Helena Gellerman
   SAFER
   helena.gellerman@chalmers.se

Applications and equipment

Applications tested

Adaptive Cruise Control, Forward Collision Warning with Emergency Brake, Lane Departure Warning, Blind Spot Information System, Electronic Stability Control and Impairment Warning (Driver Alert Control or Driver State Sensor).

Vehicle

14 vehicles (7 Volvo cars, 3 SAAB cars, 2 Volvo trucks, 2 Scania trucks), a total of 171.440 km for 2944hrs in 7934 trips, during a data collection period of over 6 months.

Test equipment

In this project, there was an initial intention to get a new (partly custom) hardware set to use as the Data Acquisition System (DAS). Due to several problems with this implementation, it was decided to go back to an off-the-shelf solution for most of the vehicle installations.

Peripherals

Eye trackers, CAN-gateways, min. 6 cameras (1 to 3 front view, 1 full internal view, 1 face, 1 rear view - only in passenger vehicles, 2 blind spots cameras - only in trucks), IR illumination, Accelerometers, Ethernet device, GPS device, Wireless communication GPRS/3G, Hard Drives.

Methodology

Method for the baseline

Depending on the vehicles, the baseline lasted from 1 to 3 months.

Techniques for measurement and data collection

Once on the road, the data was stored locally in the vehicle, while summary/status information was uploaded remotely via wireless 3G/GPRS. When it was identified that the local hard drive was about to be full, the hard drives were switched manually by OEM personnel, and the data were uploaded.

Collected data

Objective

GPS (1 Hz)

CAN (Steering Wheel Angle, Turn Indicator, Gear Level Position, Accelerations…..)

Video (Analogue) 6 Cameras in total

Extra ”external” sensors (Accelerometers, Eyetrackers – SeeingMachines/SmartEye (13units total), Lanetracker/ForwardDistVel – possibly MobilEye)

Subjective

Background questionnaire (car and truck version), Driver behaviour questionnaire (car and truck version), Decision making questionnaire, Traffic locus of Control questionnaire, User uptake/acceptance questionnaire (one per function tested, repeated at three occasions), Evaluation questionnaire (one per function tested).

Non-partners of SeMiFOT may only get access to the date through approval by SAFER and the concerned OEM.

Recruitment goals and methods

39 primary and secondary drivers (27 male, 11 female, mean age 41.3), divided between car drivers, family members to those drivers and truck drivers.

Methods for data analysis, evaluation, synthesis and conclusions

• Analysis methods:
  • Methods innovation
  • UMTRI assistance on ADAS safety impact assessment
  • Tool for accident research (MoU between SAFER and SHRP2)
  • Tool for countermeasure innovation
  • Consumer systems and insurance for large-scale data collection in future

Sources of information

SeMiFOT project reports: http://www.chalmers.se/safer/EN/publications/project-reports