1	IntelliDrive^SM Mobility Data Capture and Management Gene McHale, FHWA Office of Operations R&D TRB ITS Committee Midyear Meeting Irvine, California September 21, 2010
2	Outline Vision of Data Environments Key Issues Prototype Data Environment Sample Research Questions Future activities
3	IntelliDrive^SMMobility
4	Creating a Data Environment Data environment: well-organized collection of data of specific type and quality captured and stored at regular intervals from one or more sources systematically shared in support of one or more applications
5	Elements of Data Capture and Management Meta data: Provision of well-documented data environment Virtual warehousing: Supports access to data environment and forum for collaboration History/context: Objectives of data assembly Governance: Rules under which data environment can be accessed and procedures for resolving disputes
6	Key Issues Intellectual property rights Privacy Governance Standards and Regulation Meta-Data Quality Assurance Storage Access and Security Operations and Maintenance
7	Data Capture Prototype Data Environment https://datacapture.noblis.org/ Data (and meta-data) from the Michigan IntelliDrive Test Bed Documented probe data samples from recent tests (POC/NCAR) Trajectory Conversion Analysis (TCA) program Simulated 100% market penetration data for the test bed contributed by the University of Michigan Transportation Research Institute (UMTRI) Forums for researchers to register projects, flag erroneous data, contribute analyses and data views
8	Log Files from Onboard Equipment (OBE)
9	Snapshots Collected by Roadside Equipment (RSE)
10	Trajectory Conversion Analysis (TCA) Program Open Source Python program Inputs Vehicle Trajectory Files (Actual or Simulated) RSE locations and range Parameters for snapshot generation & transmission Outputs Snapshots generated and transmitted to RSEs Snapshots discarded for buffer overflow or RSE gap Probe segment number changes
11	UMTRI Paramics Simulation of the Testbed 189 Origin/Destinations Models 6 AM – 11 AM 10,000 concurrent vehicles RSE interactions modeled
12
13	POC OBE Data Sample Research Questions At what distance are the messages being transmitted to RSEs? Is the distribution of distances normal or uniform or constant or something else? What percentage of the snapshots appearing in the OBE log files also appear in the RSE files (i.e. were transmitted successfully)? Does the percentage vary substantially by vehicle? When transmission is unsuccessful, does the OBE attempt to send the same set of snapshots later? Does it keep trying until the transmission is successful? How often did buffer overflow occur in the OBEs and under what conditions? Use of the trajectory data to validate the speed from the snapshots
14	POC RSE Data Sample Research Questions Can the snapshots be used to estimate travel time on arterials and queuing statistics at signalized intersections? Are reported speeds consistent with link/locations? Are reported speeds consistent between different vehicles on the same link? Can an algorithm be developed to calculate the speed or travel time of a road section? What sort of “filters” would it make sense to have for access to selected subsets of RSE data? If all this data collected by RSEs is added to the data environment, how much space will be needed per probe vehicle on a daily basis?
15	Trajectory Conversion Analysis Sample Research Questions How would POC or NCAR trial snapshots have looked running the default snapshot generation process? The periodic snapshots are generated based on the speed of the vehicle. If we generated the periodic snapshot more or less frequently what would be the resulting change in snapshot content? What would be the effect in terms of snapshots received of changing the policy for PSID changes, and PSID “gaps” that prohibited the vehicles from sending to the same RSE more than once with the same PSID?
16	Paramics Simulation Data Sample Research Questions Look closely at stop/ start snapshots for single intersections; can they tell us about queue lengths and intersection overload? Are the ratios of messages received at various RSEs similar to the ratios for the actual Michigan location? Are the modeled speeds for a link close to the actual reported speeds? What is the effect of various market penetrations on the ability to extract useful traffic information from the snapshot data? At what market penetration (if any) does it become impossible to identify two sets of data labeled with different PSNs as coming from the same vehicle?
17	Future Activities Investigations into state-of-the-practice for technologies, institutional and policy, and standards Broad Agency Announcement (BAA) for Test Data Sets Acquisition and hosting of IntelliDrive funded data, such as Near term performance measures demo Safety Pilot Mobility Application Demos IntelliDrive Test Bed Development of data environments to support Mobility applications