Understanding and Shaping Urban Mobility Using Dollars and Data II
summary
Built a pipeline that ingests fare card tap data, noisy GPS and taxi traces, heals gaps and fuses sources into a digital replica that reconstructs train locations, assigns passengers to specific trains, and reports metrics (missed trains, waits over 10 minutes). Validation uses purposive ground‑truth trips and city demos; models combine LWR PDE traffic flow with an Earth‑Mover distance convexity result to optimize monetary incentives — an illustrative shift of a peak half‑hour yields a 37% reduction in modeled congestion cost.
AI-written summary · about
chapters (14)
- 0:00Intro
- 0:58Overview of Presentation
- 7:53Digital Replica of Metro System From Ticketing Data
- 22:07Pickups from Midtown Manhattan
- 23:38Pickups from the Financial District
- 23:49Implication
- 28:56Tipping Percentage
- 37:16The Highway Model
- 38:54Fluid model of highways: The LWR model
- 41:20Density-dependent velocity
- 42:00Solving the PDE
- 44:30Approximation to the highway map
- 45:53The cost of congestion
- 56:08(An unrealistic) Example
related talks
29mIncentives and Games in Large Scale Mobility
48mUrban Data & Decisions
34mData Drives Business
35mExposing Algorithms
21mExpert Panel | Going Places: Location Intelligence for Transport and Logistics
10mCivic Hacking & Urban Informatics: Mapping Data with GeoPandas
13mCites Digital Transition to a Sustainable Future: The Role of Data
19mThe Data Science behind Cruise’s Robotaxi Fleet