Reconstructing Transportation Network Dynamics: Big Data Systems, Algorithms and Visualization
Big Data Series
Professor Balaji Prabhakar
The transportation networks of cities are large-scale and exhibit complex dynamics. For example, the daily ridership of public transport systems is 3–10 million trips with an average of 12–20 kms per trip, and the stochastic process governing the dynamics of the network records the spatio-temporal movement of these large numbers of commuters and a large number of fleet vehicles. Of interest to transportation network operators and planners are measures of performance such as delays due to congestion, wait times at stations, measures of crowdedness, and fleet-level details. These quantities are not directly observable. The observations that are available are sparse and, often times, aggregated. The following is a question of significance in this setting: Can the stochastic process corresponding to the transportation network be reconstructed from the sparse observations?
In this talk, I will present algorithms for reconstructing road, train and bus networks; the big data systems needed to crunch the huge volume of data; and some natural visualization frameworks. I will also describe learning systems that can understand commuter behavior with changes in the weather, lane closures and traffic accidents, etc.