90 likes | 195 Views
Generating Summaries from FOT Data ITS World Congress, Detroit 2014 Dr. Sami Koskinen, VTT sami.koskinen@vtt.fi. Data Processing in DRIVE C2X and TeleFOT. Field Operational Tests (FOTs) are large-scale user tests which aim at comprehensive assessment as well as promotion of latest functions
E N D
Generating Summaries from FOT Data ITS World Congress, Detroit 2014 Dr. Sami Koskinen, VTT sami.koskinen@vtt.fi
Data Processing in DRIVE C2X and TeleFOT • Field Operational Tests (FOTs) are large-scale user tests which aim at comprehensive assessment as well as promotion of latest functions • FOTs collect detailed data for assessment, commonly in the range of terabytes. Data comes from various sensors, traffic and weather information systems, communication, functions etc. • This presentation covers data processing approaches used in EU projects DRIVE C2X and TeleFOT, where multiple FOTs were carried out • Data was shared between partners, documented in detail and analysed collaboratively across test sites. • Analyses concerned safety, traffic efficiency, environment, user acceptance and technical performance
Motivation for Generating Summaries • In FOTs, the amount of collected data is generally too large for the test to be comprehensively analyzed without first generating summaries. • Data can be split over multiple hard drives and each simple calculation may take a week to complete. • Driving diaries and event lists are common summary tables • When several analysts collaboratively work with FOT data, same post-processing and set of indicators make their work more efficient • A couple of professional programmers can implement hundreds of indicators based on analysts’ requirements • Summarized content minimally gives an index into raw data.
Harmonisation Across Tests • Harmonised map matching and calculation of indicators and summary tables reduces individual analyst’s work and also ensures comparability of indicators across tests • Post-processing handles different log file formats, function trigger and communication definitions, lists of broken loggers and important dates such as changing to winter-time speed limits • As a result of post-processing, analysts get similar summary data sets from each FOT • Variables derived from logger data are calculated using the same definition • Shared post-processing also helps to avoid some coding errors that would happen if each test site / analyst works separately
Main Summary Tables - Legs • Driving diaries, where each leg is described by a long list of derived variables such as time stamps, total distance driven, number of hard braking events, fuel consumption and percent driven on a road type • More than 200 derived variables / indicators for statistical analyses • One caninstantlygeneratereportsfrom the summarytables, e.g.
Main Summary Tables - Events • Table rows describing events, e.g. periods of HMI activity: coordinates, information shown, speed at the beginning and end • Can also be lists of e.g. speeding events • Enables selection for analysis GPS Visualizer & Google
Conclusions • Harmonizing logging and post-processing enables analysts to more easily cover several tests in collaborative projects. • Summarized content minimally gives an index into raw data. In the best case, analysts work together with professional programmers, providing them the data for further analyses. • Summary tables are of manageable size whereas the size of raw FOT datasets often causes practical problems • Enriching and combining different data is beneficial: map data linked with coordinates, events linked with manual video annotations and driving diaries with user and vehicle data. • Data documentation and sharing are the keys for comprehensive analysis!