Traveler Centered Studies Decentralizing Data & Benefiting Travelers Dr. Ross Gore Virginia, Modeling, Analysis & Simulation Center
Standard Practice
Side Effects Personnel is paid to either install a device or observe and collect data Error prone Expensive Very discrete Data is centralized Storage and security concerns Participants are unaware they are involved Traveler-specific decisions cannot be made
Why do we centralize data? Reason is largely analytical Test hypotheses via modeling and simulation Really just need intermediate results that can be aggregated
What if …. ? Smartphones & On-board devices collected data computed intermediate results Privacy can be maintained Storage of data isn’t an issue Traveler specific analysis is possible
Example: Fitting Statistical Models
Data Collection terminates as soon as fit occurs!
Is this realistic ? Simulated fitting a logistic transport model Individuals opt into a study at a fixed rate Repeated measurements occur with a fixed amount of time between them 90% probability of individual having their smart phone on at any particular moment 90% probability of individual traveling
Is this realistic ? Simulated fitting a logistic transport model
What’s in it for the Traveler A DriveSense App Store is needed to centralize experiment apps for drivers to download Publish finished models in the App Store for drivers to use. Participate in a study receive a download credit Model is traveler specific because it uses data the traveler is currently collecting!
Total Vision
Conclusion A distributed, traveler-centric approach has benefits No data privacy, storage or maintenance issues Previously collected data reduces burden on new studies Studies can conclude as soon as sufficient data is collected Resulting models/tools can use previous & newly collected traveler data for custom results Participation is encouraged by giving access to resulting tools