 |
 |
 |
 |
Data Acquisition & Field Test Data Analysis
Topics
- Overview - purpose of data collection
- Data archive overview
- Requirements on data
- Data acquisition
- Fleet monitoring
Data Acquisition System
- Phase I
- Support Phase I development of IVBSS.
- Provide data during Phase I testing.
- Phase II
- Provide data from extended pilot FOT and from FOT itself.
- Allow remote monitoring of test fleet (IVBSS performance & health, driving behavior )
Data Uses
- UMTRI–
- monitoring fleet
- analysis of experimental data (performance, acceptance, safety)
- debriefing test subjects
- Visteon/Eaton/Cognex –
- Volpe Center/USDOT–
- quality assurance
- analysis of experimental data
- Side benefit – data archive for future research
Data Archive – “Raw” Objective Data
-
Driver information
- Onboard data
- Numerical (database)
- Video
- Audio
- Offboard data
- Map & roadway feature databases
- Weather data
Data Archive – Processed Objective Data
- Cleansed data – smoothing/fusing, managing dropouts & outliers, bias removal, validating trips, etc.
- Driving context characterization using roadway data, weather data, time, etc.
- Characterization of events & scenarios of interest
- Building measures of system performance, potential safety impacts, facts possibly affecting driver acceptance
Additional data may exceed size of original raw numerical data
Data Archive - Subjective Data
- Pre-drive questionnaire
- Driving style and behavior questionnaires completed prior to FOT participation
- Post-drive questionnaires
- Completed after FOT participation: on-site and take home
- Driver debriefs
- Review of a subset of warnings to rate for usefulness
- Focus groups
Data Archive - Subjective Data: Questionnaires
- Pre-drive questionnaires
- Driver style questionnaire (DSQ)
- Evaluates 6 factors of drivers' style: focus, calmness, social resistance, speed, deviance, and planning
- Driver behavior questionnaire (DBQ)
- Examines drivers' errors, lapses, and violations
- Drivers' scores from DSQ and DBQ will be used in statistical models as predictors of IVBSS acceptance
- Post-drive questionnaires
- Extensive evaluation of drivers' opinions of IVBSS
- Will evaluate safety, ease of use, comfort and convenience, and willingness to purchase
- Two questionnaires
- On-site: highest priority questions
- Take home: questions of lesser priority
Requirements on Data Archive
- Complete and ‘auditable’ characterization of events & system performance
- Highly robust & structured data set
- Continuous 10 to100 Hz logging (depending on subsystem)
- ~400 signals on light vehicle and ~300 signals on heavy truck
- Video collection to provide analysts with situational context for FOT data, especially IVBSS-related events.
- What was happening inside and outside the vehicle?
- What did the IVBSS system react to?
- Secure from data loss, privacy concerns
Data Archive Formats & Size
- Format when archived:
- Numerical: enterprise-level relational database
- Video: MPEG-4 compressed video with indexing for synching with numerical data
- Compression & frame rates vary by video stream
- Audio: Compressed 64 kbps with indexing
- Size estimates are preliminary
- Numerical – depends on radar – 1-2 terabyte (TB) order of magnitude
- Video – depends on compression levels – 10 TB?
Requirements on Data Archive
- Highly usable:
- Analyst access to all data within seconds, including video
- Analysis tools
- Sharing information between project team & independent evaluator
Data Acquisition System
- Two CPU system (CAN/radar + vision/audio)
- Automotive-grade hard disks
- CAN and J1939 buses - primary data sources
- Second GPS for analysis (differential)
- 5 cameras with video capture & compression
- Up to 7 radars
- Vehicle motion sensors
- GPRS/Edge cellular modem
- DAS power management system
Monitoring the fleet
Track:
- health of IVBSS & data system
- usage of vehicle
- driver experience with IVBSS (alert types & experience)
Cellular modem:
- Trip characteristics, IVBSS actions, health information including histograms
Summary
- Data archive – several types of data
- Extensions from previous FOT analyses:
- More advanced data collection
- Power through joining diverse types of data: onboard, offboard, driver information.
- Driving treated as more ‘holistic’ than previous studies – context considered in more detail
|
 |
 |
 |
 |