Follow Datanami:
December 3, 2020

‘TranSEC’ Machine Learning Tool Tackles Los Angeles Traffic

Oliver Peckham

Traffic management is essentially a giant optimization problem, with urban planners and traffic engineers aiming to reshape roads, traffic signals and parking to smooth traffic patterns and avoid serious congestion. However, there is a major obstacle in this quest: data availability. To date, most traffic engineers rely on spotty, incomplete traffic and collision data based on infrequent samples. Now, Pacific Northwest National Laboratory (PNNL) has developed a new tool, TranSEC, that aims to produce more holistic data for traffic optimization.

TranSEC leans on a dataset that has been under development for years: Uber ride data. Using data from Uber rides, combined with traffic sensor information, TranSEC map uses machine learning to map traffic flow over time. The machine learning elements of the tool extrapolate from the incomplete Uber data to generate a near-real-time picture of traffic. 

The TranSEC team began with data from the Los Angeles metropolitan area (around 1,500 square miles), managing to create a traffic congestion model in minutes where it would have taken hours using previous tools.

“What’s novel here is the street level estimation over a large metropolitan area,” said Arif Khan, a PNNL computer scientist who helped develop TranSEC. “And unlike other models that only work in one specific metro area, our tool is portable and can be applied to any urban area where aggregated traffic data is available.”


Video courtesy of Graham Bourque | Pacific Northwest National Laboratory.

For now, the researchers are only using TranSEC to develop congestion models, but as the machine learning components of the model ingest more information, its predictive capacity will eventually become robust enough to serve in the development of corrective measures.

“We use a graph-based model together with novel sampling methods and optimization engines to learn both the travel times and the routes,” said Arun Sathanur, a PNNL computer scientist who helps lead the TranSEC research team. “The method has significant potential to be expanded to other modes of transportation, such as transit and freight traffic. As an analytic tool, it is capable of investigating how a traffic condition spreads.”

For now – at least, for Los Angeles – TranSEC generally requires the sorts of high-performance computing resources found at major national laboratories like PNNL. However, TranSEC is scalable, and PNNL claims that the tool can be run at a lower resolution (just major roads) on a single high-end desktop computer.

“Traffic engineers nationwide have not had a tool to give them anywhere near real-time estimation of transportation network states,” said Robert Rallo, principal investigator for the TranSEC project. “Being able to predict conditions an hour or more ahead would be very valuable, to know where the blockages are going to be.”

Datanami