Communications of the ACM
Building a Better Traffic Forecasting Model
Argonne National Laboratory scientists are using artificial intelligence to forecast large-scale traffic patterns.
Credit: Bus & Motorcoach News
Scientists at the U.S. Department of Energy's Argonne National Laboratory (ANL) are using artificial intelligence to forecast large-scale traffic patterns with greater accuracy.
The research team fed ANL supercomputers traffic patterns from roughly a year's worth of data culled from 11,160 sensors along the California highway system.
The team trained a machine learning model on that data to forecast traffic within milliseconds; the model can analyze the past hour of data, and anticipate the next hour of traffic.
The ANL researchers dramatically shortened the time required to train the model, from a week on a top-of-the-line desktop to three hours on a supercomputer.
The model applies graph-based deep learning to predict traffic patterns from historical patterns, while concurrently forecasting traffic speed and flow.
From Argonne National Laboratory
View Full Article
Abstracts Copyright © 2020 SmithBucklin, Washington, DC, USA
No entries found