Data Analysis on Nomadic Systems: The TeleFOT Approach

Size: px

Start display at page:

Download "Data Analysis on Nomadic Systems: The TeleFOT Approach"

Loreen Collins
5 years ago
Views:

1 Data Analysis on Nomadic Systems: The TeleFOT Approach Presentation to FOT-NET Seminar, London, UK October 2010 Andrew Morris Ruth Welsh Loughborough University, UK.

TeleFOT TELEFOT TeleFOT - Field Operational Trials of Aftermarket and Nomadic Devices in Vehicles. Runs from June 2008 to May 2012.

2 TeleFOT TELEFOT TeleFOT - Field Operational Trials of Aftermarket and Nomadic Devices in Vehicles. Runs from June 2008 to May Funded by the EC and National Administrations budget is 15million Project partners (22 in total) include research institutes telecomms, manufacturers and software companies

3 TeleFOT Field Operational Tests Operational Area Co-ordinated by VTT (Finland) 3 test Communities (Northern, Central and Southern Europe) Approximately 3,000 drivers Test-sites in Finland, Sweden, UK, Germany, France, Italy, Spain and Greece

4 Functions and TELEFOT Systems Functions to be to tested be (examples) Traffic and travel Information systems Speed limit information, driver warning and feedback Personalised Navigation Devices (PNDs). Forward Collision Warning/Lane-Keeping Systems ecall Green driving advisory systems All of the above are available on an aftermarket device such as a smart-phone or other device

5 Working areas in TeleFOT TeleFOT 2 PY review meeting 23/1 5 0/20

6 A Field Operational Test TeleFOT Impact Areas Research questions and hypotheses must cover all TeleFOT impact areas: Efficiency Environment Mobility Safety User up-take (Acceptance and adoption) 6

7 The TeleFOT state of play Problems Policy objectives Impacts wanted Problems solved? Policy objectives met? Impacts IRL? Function identification & description The FOT Chain Socio Economic impact assessment Use Cases Research Questions &Hypotheses Ethical and legal issues System & Function Analysis Research Questions and Hypotheses analysis Implementation Plan Performance indicators Measures & Sensors Study design Data acquisition Data decoding Measures Database Data Analysis Performance indicators 7

8 The TeleFOT state of play (continued) The general methodology of TeleFOT has been established in terms of; Devices and functions to be tested Research questions and hypotheses to be addressed The associated Determination of performance indicators Data-logging Subjective data collection tools (questionnaires etc) The Pilot tests are now running at about half of the test-sites The main FOTs will start later this year (November/December) Pilot analyses are under way to test the end-to-end data flow Stakeholder consultations are underway 8

9 Integration of top-down & bottom-up approaches Research Questions Top Down approach PREVIOUS STEP Theoretical basis for impacts PREVIOUS STEP Functions Types of impact Generic Research Questions 1. Function 2. Device 3. Use case 5. Research Hypothese s Performance indicators Study design Measures and sensors Devices Use cases 4. Types of impact Bottom Up approach Iterative process: Cost/benefit analysis of investigating each hypotheses 9

10 How to select a final set of hypotheses from the wish list A trade-off of COST against BENEFIT COST = measures, loggers, data collection time, analysis time BENEFIT = the importance of understanding the implications (for safety, mobility, efficiency, environment, user uptake) An iterative process to be undertaken by: the project as a whole the individual test sites The feasibility of addressing the hypothesis within the FOT 10

Mobility impact Top-down approach Amount of travel number of journeys length of journeys Travel patterns travel mode route time budget timing

11 Mobility impact Top-down approach Amount of travel number of journeys length of journeys Travel patterns travel mode route time budget timing destination Mobility choice mobility choice enablers mobility choice constraints Journey quality stress subjective safety comfort other quality issues 11

Navigation hypotheses bottom-up MOBILITY (+) Result improved mobility due to route choices more efficient use of road capacity (+) Save time for pre-trip planning (+) Will increase number of

12 Navigation hypotheses bottom-up MOBILITY (+) Result improved mobility due to route choices more efficient use of road capacity (+) Save time for pre-trip planning (+) Will increase number of journeys/exposure (-) Inability to plan routes without the system SAFETY (+) Will result in more use of major roads (+)Will result less time/exposure (-) Will result in increase in mean speed (-) In long term use will result delegation of responsibility (-) Will increase number of journeys/exposure (-) Will increase visual demand and distraction ENVIRONMENT (+) Will result less pollution and noise (+) Will result less pollution and noise in residential area (-) Will result in increase in mean speed, consequently increase in fuel consumption, noise and emissions (-) Will increase number of journeys/exposure 12

TeleFOT Database System Architecture FOT Framework results Impact

collection specifications Logged Data analysis Performance

designs Combined Pre-processed Database (MySQL) Result file Example

(MATLAB, SPSS) Impact area data analysis methods Off-board services

13 TeleFOT Database System Architecture FOT Framework results Impact area Research Questions Vehicle DAS Local DB Hypotheses Data collection specifications Logged Data analysis Performance indicators EMTELE FILTERING (according to journey or link) Study designs Combined Pre-processed Database (MySQL) Result file Example SQL-requests Data request (SQL) File import in analysis program (MATLAB, SPSS) Impact area data analysis methods Off-board services Service logs Broadcast data log Questionnaires Map matching Scanned documents Weather info 13

14 Data Quality checks within TeleFOT Data Quality is an essential feature of TeleFOT A theoretical data-flow chart is available within the project test-sites are expected to conform to the check-list to ensure data quality and consistency at the individual stages Data collection (in normal driving situations) Data download (from the vehicles) Data transfer (to the back-office) Data cleaning Data upload (to central server) Data download (from CS to test-site) Data analysis 14

15 TeleFOT - Impact Assessments Analyses will employ a variety of statistical and non statistical techniques to answer the Hypotheses Both quantitative and qualitative data will be used Requirement for additional questionnaires / interviews /travel diaries to be undertaken in order to explain any effects seen Assessments require some analyses of video data Pilot analyses are now been carried out to check Data-flow in the project That the data address the Research Questions 15

TeleFOT Data Analysis Types of Data Core Objective Data (from Data Loggers) Subjective Data (from Questionnaires and travel diaries) Subjective Data (from Focus Groups) Test-site-specific Data

16 TeleFOT Data Analysis Types of Data Core Objective Data (from Data Loggers) Subjective Data (from Questionnaires and travel diaries) Subjective Data (from Focus Groups) Test-site-specific Data collected in Detailed FOTs (video-data, enhanced data-loggers, etc) Simulation/modelling data based upon patterns observed in the logged data All above forms of data will be used to answer the specific Research Questions and associated hypotheses listed in the Analysis Plans 16

17 Core Data from Data Loggers - Example Field Description Length Example Interpretation 1 GPS Data Minimum, Record ID for GPS data. 3 GDM 2 UTC time: Hours, minutes and seconds :33:37 3 Latitude coordinates in the degrees/minutes format ' 4 North (N) or south (S) side of the equator 1 N North 5 Longitude coordinates in the degrees/minutes format ' 6 East (E) or west (W) side of the Prime Meridian. 1 E East 7 Altitude meters above mean sea level m above sea level 8 Elevation in degrees, 90 maximum Azimuth, degrees from true north, 000 to Speed, km/h (what is the measured direction of the speed) km/h 11 Number of satellites being tracked satellites being tracked 17

18 Parametric Analyses (from logged data) Mostly related to Large-scale FOTs Use of logged data including GPS, accelerations, braking incidences, number of trips Statistical analysis of data relating to trips Baseline and experimental conditions 18

TeleFOT LFOT Hypothesis Testing Example Data Level The research hypothesis The impact assessment area Example The Device changes the number of trips undertaken by the subjects Safety/User uptake FOT

19 TeleFOT LFOT Hypothesis Testing Example Data Level The research hypothesis The impact assessment area Example The Device changes the number of trips undertaken by the subjects Safety/User uptake FOT Type The Devices to be evaluated Large-Scale FOT Navigation support and Traffic Information The Analysis The statistical tests Subjects response; - More because of device - Less because of device - Device had no influence on number of trips Multiple regression, Odds ratio, ANOVA, as appropriate 19

Non-Parametric Analyses Specific to Detailed FOTs Use of in-depth data collection procedures Higher accuracy acquisition capability Multi-channel data logging Video footage Specialist sensors

20 Non-Parametric Analyses Specific to Detailed FOTs Use of in-depth data collection procedures Higher accuracy acquisition capability Multi-channel data logging Video footage Specialist sensors such as Eye-tracking/Cockpit Assessment Activity Analysis of Glance Behaviour and EORT Allows an experimental approach to be implemented which in turn allows statistical modelling to be applied 20

21 Safety Top level Research questions RQ1 Is the route affected (where travel takes place)? LFOT / DFOT RQ2 Is the amount of time on the road affected (how long travel takes place for)? LFOT / DFOT RQ3 Does the device cause distraction? DFOT RQ4 Is speed affected? LFOT RQ5 Is vehicle positioning affected (proximity and lane positioning)? DFOT RQ6 Is braking affected? LFOT / DFOT RQ7 Is non driving manual activity affected? DFOT

DFOT Hypothesis Testing Example Data Level The research hypothesis The impact assessment area FOT Type The Devices to be evaluated The Analysis The statistical tests Example The Device causes

22 DFOT Hypothesis Testing Example Data Level The research hypothesis The impact assessment area FOT Type The Devices to be evaluated The Analysis The statistical tests Example The Device causes distraction (SRQ-5) Safety Detailed FOT All Devices Eyes off road time Glance behaviour Workload (CAA) Safety-Critical Events (Decelerations, Harsh-braking, Yaw-rates) related to device Distance to Lead vehicle, TTC, Long/Lat positioning etc (ECA) Mainly descriptive statistics Statistical modelling if required 22

23 11/10/2010 TeleFOT1st Year review, Brussels, 10 th and 11 th September

24 Visual Data Analysis Example of Assessment 24

25 FaceLAB visualisations The FaceLAB system can be configured to create a 3D World from the data Ford Mondeo Interior 3D world 25

data Red Line = Head pose Green Line = Gaze

world view (green highlight) GAZE_OBJ_NAME

734239-0.0388254-0.05 0.734239-0.238825 nothing 0 0.

673476-0.0345702-0.05 0.673476-0.23457 nothing 0 0.

181694 RH_M 0 0.639698 0.0726293-0.05 0.

05 0.540895-0.0876678 nothing 0 0.479261 0.

26 FaceLAB visualisations 3D world Mondeo Interior Addition of logged eye and face data Red Line = Head pose Green Line = Gaze direction Intersection shows a glance in world view (green highlight) GAZE_OBJ_NAME GAZ_SCR_INTSECT GSI_WORLD_X GSI_WORLD_Y GSI_WORLD_Z GSI_SCREEN_X GSI_SCREEN_Y nothing nothing nothing nothing nothing RH_M RH_M nothing nothing nothing Data shows a glance from Nothing (the outside view) to the RH_M (right hand mirror) and back. 26

27 FaceLAB visualisations 3D world Another interaction Rev counter 27

28 FaceLAB visualisations Fully navigable world... Drivers eye view 28

29 Visual behaviour variables (per period of analysis) Number of glances to each area Duration of each glance to each area % time looking at each area 29

30 Glance Duration with and without System 30

31 Glance behaviour with and without System 31

32 Challenges using the FESTA Guidelines Ideal Flow (based on FESTA) Research Hypotheses Research Questions Performance Indicators Data Architecture Data Quality Data Handling Data Procedures Data Requirements Data Logging Transmission & Management Data Sources Identified Data Specifications Data Log Specifications Data Loggers Spec 32

33 Challenges within TeleFOT in the context of FESTA FESTA is a useful check-list for the whole project and adherence to the elements within the chain is appropriate Using the FESTA Protocol in a multi-partner collaboration Partners make progress at different rates and keeping to a sequential time-plan is difficult Different partners have different priorities in the sequence Dealing with combination of functions this is also common to ADAS functions! 33

34 Challenges within TeleFOT in the context of FESTA Analysing in the context of use-cases will be difficult and sometimes irrelevant Examples in the FESTA guidelines are country specific - i.e. This is the case in Germany but it may be different in...(may be better to have as a points to consider ) 34

Contact us Andrew Morris, TeleFOT SP4 Leader Loughborough University Tel: +44 1509 226981 Mob: 07974 824476 Fax: +44 1509 226960 Email: a.p.morris@lboro.ac.uk Mail: Vehicle Safety Research Centre ESRI, Loughborough University Garendon Building, Holywell Park, Loughborough, Leics, LE11 3TU, UK More info on: www.

35 Contact us Andrew Morris, TeleFOT SP4 Leader Loughborough University Tel: Mob: Fax: Mail: Vehicle Safety Research Centre ESRI, Loughborough University Garendon Building, Holywell Park, Loughborough, Leics, LE11 3TU, UK More info on:

TeleFOT, eld operational tests of aftermarket nomadic devices in vehicles, early results

Loughborough University Institutional Repository TeleFOT, eld operational tests of aftermarket nomadic devices in vehicles, early results This item was submitted to Loughborough University's Institutional