Karl Rehrl, R. Brunauer, M. Hufnagl, S. Leitinger, A. Wagner and M. Wimmer Floating Car Data Testbed Salzburg Motivation, Goals and First Results
Outline Motivation and Goals Data Collection, Processing and Analysis Data Privacy Fleet Characteristics Extended Floating Car Data Future Plans 23.10.2013 Karl Rehrl, Mobile Ghent '13 2
Motivation Floating car data technology is in widespread use Taxi FCD (Brockfeld, Wagner, Lorkowski, & Mieth, 2007) TomTom, Inrix, Google, Car manufacturers Lack of independent and scientifically motivated FCD testbeds Stockholm Case Study (Mahmood, Koutsopoulos, & Ranganathan, 2010) Results from wide-area field trials with different fleets are missing Results for rural areas are missing Partially-controlled experiment: Possibility to ground-truth data 23.10.2013 Karl Rehrl, Mobile Ghent '13 3
Floating Car Data Testbed Salzburg Province of Salzburg Area: 7.154 km 2 Population: 530.000 140 km motorways 1.400 km primary and secondary roads City of Salzburg Area: 65,68 km 2 Population of 150.000 40.000 daily commuters Source: SAGIS, Land Salzburg 23.10.2013 Karl Rehrl, Mobile Ghent '13 4
Goals of the FCD Testbed Salzburg Test of different technologies for near real-time GPS data collection Built in GPS devices, Plug-in GPS devices, Smartphones Different GPS sampling rates: From 1 second to 60 seconds Different fleets: commercial fleets, private cars, public transport fleets Highly scalable floating car data system based on open source software Designed to process GPS data from thousands of cars in near real-time Works with GPS trajectories All trajectories are stored for historical analysis High standards concerning data privacy Full accordance to European data protection laws 23.10.2013 Karl Rehrl, Mobile Ghent '13 5
Data Collection Different Approaches GPS devices Data brokers Smartphones Falcom Bolero (FCD) Falcom FOX-EN (XFCD) Smartphone App Push Vehicles AMV ASG (XFCD) Push Aplicom A9 (XFCD) Push /Pull 3-Party- API Data adapter Data adapter Data adapter FCD Collector Instance FCD Collector Instance FCD Collector Instance... GPS Messages FCD / X-FCD 23.10.2013 Karl Rehrl, Mobile Ghent '13 6
Data Collection with Smartphones Tests are done with an Android-based application + Smartphones are well-suited for GPS data collection + No additional equipment of vehicles with GPS devices necessary - Users have to actively start the application - Users have to be motivated to start the application and keep it running - Battery consumption is still an important issue Collected data is only useful with reliable mode detection Brunauer, Hufnagl, Rehrl and Wagner (2013): Motion Pattern Analysis Enabling Accurate Travel Mode Detection from GPS Data Only, In: Proceedings of the 16th International IEEE Annual Conference on Intelligent Transportation Systems, The Hague, The Netherlands 23.10.2013 Karl Rehrl, Mobile Ghent '13 7
Data Processing Overall Workflow GPS Data API Collect Data Assemble Tracks Track Messages Track Storage GPS Messages Anonymize Tracks Analyze Tracks Track Database Street Graph Travel Times & LOS Individual Analysis 23.10.2013 Karl Rehrl, Mobile Ghent '13 8
Data Analysis Near Real-time & Historical Data Level-of-Service (LOS) Calculation Reference Speed Calculation near real-time, re-calculation every minute reference speed calculation for all street segments, 80% quantile of all travels, update every night 23.10.2013 Karl Rehrl, Mobile Ghent '13 9
Data Privacy General Principles High standards for data privacy In accordance with Austrian as well as European data protection laws (EU-RL 95/58/EG approved on 31.10.2003) Fleets: Commitment of staff association Private persons: Individual commitments All trajectories are collected and stored anonymously Avoiding direct links to individual persons -> random numbers for trajectories Avoiding direct links to individual vehicles -> random numbers for trajectories, splitting Avoiding indirect links to persons -> splitting trajectories, cutting trajectories at start and end 23.10.2013 Karl Rehrl, Mobile Ghent '13 10
Data Privacy Cutting of Trajectories 1 3 2 4 S..Street segments TS..Trajectory start TE..Trajectory end 5 23.10.2013 Karl Rehrl, Mobile Ghent '13 11
Fleets Different Characteristics Fleet Operator # Vehicles km / month km / car / month Mobile health care 116 Ø 88.000 km Ø 760 km Ambulance 150 Ø 190.000 km Ø 1.266 km Breakdown service 16 Ø 40.000 km Ø 2.500 km Car distributor 25 Ø 40.000 km Ø 1.400 km Taxi services 26 Ø 55.000 km Ø 2.115 km Utility company* 500 n.a. n.a. Private persons 26 Ø 22.000 km Ø 850 km City busses 100 Ø 400.000 km Ø 4.000 km Regional busses* 100 n.a. n.a. * planned 23.10.2013 Karl Rehrl, Mobile Ghent '13 12
Fleets Regional Coverage (one month of data) ambulance car distributor breakdown service taxi services private cars mobile health care 23.10.2013 Karl Rehrl, Mobile Ghent '13 13
Extended Floating Car Data (XFCD) Data from vehicle on-board electronics Data from different vehicle sensors may be collected GPS devices have to be connected to the CAN bus (Convenient CAN) Messages on the CAN bus vary between car models Values are stored together with GPS trackpoints <trkpt lat="47.691483" lon="13.109062"> <ele>448.1000061035156</ele> <time>2013-10-20t13:16:37+0200</time> <hdop>1.5</hdop> <vdop>5.0</vdop> <sat>7</sat> <extensions name= xfcd"> <tag key="xspeed" value="68"/> <tag key="xespd" value="2221"/> <tag key="xfcon" value="104192.0"/> <tag key="xatmp" value="19.0"/> <tag key="xdind" value="0"/> <tag key="xbrlt" value="false"/> </extensions> </trkpt> 23.10.2013 Karl Rehrl, Mobile Ghent '13 14
Extended Floating Car Data - Parameters Param Description Units VW Caddy VW Golf V Skoda Fabia xatmp ambient air temperature degree Celsius + + + xfcon fuel consumption consumed fuel in µl + + - xespd engine speed rounds per minute (rpm) + + + xspeed speed km/h + + + xdibm dimmed beam headlights on/off + + + xfubm full beam headlights on/off + + + xffl front fog light on/off - - - xrfl rear fog light on/off + + + xdind direction indicator off/left/right + + + xwrlt warning lights on/off + + + xbrlt brake lights on/off + + + xwiper wiper off/interval/on + + + xrain ambient humidity sensor relative humidity - - - 23.10.2013 Karl Rehrl, Mobile Ghent '13 15
XFCD Example: Fuel Consumption (µl / meter) 23.10.2013 Karl Rehrl, Mobile Ghent '13 16
XFCD Example: Ambient Air Temperature (Celsius) 23.10.2013 Karl Rehrl, Mobile Ghent '13 17
XFCD Example: Wiper State (Off/Interval/On) 23.10.2013 Karl Rehrl, Mobile Ghent '13 18
Future Plans FCD Testbed Salzburg is fully operational since May 2013 Current daily data volume: 3.500 tracks, 700.000 points, 35.000 km, 1.500 hours Continuous integration of additional fleets (several 100 vehicles) Detailed data analysis, validation of methods and algorithms Using FCD and XFCD in scientific studies, e.g. supporting PhDs, master thesis Smartphones as Sensors (ongoing master thesis) Acceleration, Magnetic Field, Illumination, Proximity, Gyroscope, Air Pressure, Relative Humidity, Ambient Air Temperature 23.10.2013 Karl Rehrl, Mobile Ghent '13 19
Any questions or comments? Dr. Karl Rehrl Head of Mobile and Web-based Information Systems Salzburg Research Forschungsgesellschaft mbh Jakob Haringer Straße 5/3 5020 Salzburg, Austria T +43.662.2288-416 F -222 karl.rehrl@salzburgresearch.at Funding bodies