A machine learning algorithm is being developed to predict traffic flow and traffic patterns on California highways.
A team of Lawrence Berkeley National Laboratory computer scientists and the California Department of Transportation, or Caltrans, are collaborating to use high-performance computing and machine learning to help Caltrans make decisions when accidents occur.
The research was done in conjunction with California Partners for Advanced Transit and Highways, or PATH — which is part of the UC Berkeley Institute of Transportation Studies — and Connected Corridors, a collaborative program for managing transportation corridors in California.
Caltrans and Connected Corridors are now implementing the system on a trial basis in Los Angeles County through the I-210 Pilot. The researchers want to coordinate multi-jurisdictional traffic incident response plans to limit the negative impacts of accidents.
Co-project researcher Hongyuan Zhan said in an email that the research is for the betterment of society because of its applicability to the real world.
“From my point of view, this work is an example of using advances in machine learning for social good,” Zhan said in an email. “In addition, the collaboration between machine learning researchers and transportation domain experts are very valuable for designing algorithms that meet with real-world data and operation constraints.”
Improving traffic conditions requires the collaborative effort of all partner agencies, according to project contributor Brian Peterson. The effort involves the political, physical, data and technical resources of each partner agency to achieve such goals, Peterson added.
Berkeley Lab senior scientist and co-researcher Sherry Li said in a Berkeley Lab press release that multiple traffic models are recommended for prediction accuracy.
“Many traffic-flow prediction methods exist, and each can be advantageous in the right situation,” Li said in the release. “To alleviate the pain of relying on human operators who sometimes blindly trust one particular model, our goal was to integrate multiple models that produce more stable and accurate traffic predictions.”
Using data collected from Caltrans sensors on California highways, the project algorithms achieved accurate predictions on a 15-minute rolling basis, according to the press release. The project team validated and integrated the algorithms using real-time traffic data the Connected Corridors system collected. The system generates predicted traffic flows at sensor points along the freeway, specifically at freeway entrances and exits.
Co-researcher Gabriel Gomes said in an email that Zhan and Peterson worked on the project over the summer in Berkeley and have built a data hub for Connected Corridors. The algorithm used was “perhaps the first” machine learning tool implemented in this data hub.
“These results are important because they represent a deployment of cutting edge machine learning algorithms in a real-world computational and data environment,” Gomes said in the email. “The work stands apart from other academic exercises where things are built in an idealized sand box.”