Deep learning to escape deep water

Posted by
On August 2, 2019

Missouri S&T engineers are using a type of artificial intelligence that mimics how the human brain makes decisions to determine the best evacuation routes during a flood. Photo by Carolina Hidalgo, St. Louis Public Radio

Artificial intelligence (AI) may soon help transportation agencies and first responders determine the best evacuation routes during floods, thanks to the work of Missouri S&T researchers.

Suzanna Long, Hist’84, Phys’84, MS EMgt’04, PhD EMgt’07, and Steve Corns, both of Missouri S&T’s engineering management and systems engineering department (EMSE), are using a form of AI known as “deep learning” to develop forecasting tools to integrate water level rate of change as part of evacuation route planning in flood-prone areas.

Deep learning is a type of machine learning that imitates the human brain’s ability to process information and create patterns for use in decision making. Long and Corns will design their deep learning model to determine the best evacuation routes based on flood data, available roads for evacuation and traffic patterns.

Supported by funding from the Missouri Department of Transportation (MoDOT) and the Mid-America Transportation Center (MATC), a U.S. Department of Transportation initiative based at the University of Nebraska-Lincoln, Long and Corns are also using geospatial data from the U.S. Geological Survey, the National Weather Service and other public data sources to build their forecasting model.

For this research project, Long and Corns will use data from the Meramec and Missouri river basins to train a deep learning neural network to determine how deep and how quickly floodwaters will rise.

“We’ll use river rise and current flood plain modeling efforts from partners as part of a deep learning model to develop algorithms to determine when and where traffic needs rerouted,” says Long, chair and professor of EMSE. “We’ll use the ground truth from the spring 2019 flooding to guide our solutions and as model inputs.”

“This will be used to find relationships between the available data to increase the overall accuracy of the deep learning neural network,” says Corns, an associate professor of EMSE.

“This rate of rise is used to model evacuation or detour planning modules that can be implemented to assure the safety of the community and highway personnel, as well as the safe and secure transport of goods along public roadways,” Long adds.

The pair will also create a routing algorithm to guide evacuations based on an assessment of available roads and their conditions. Here, the AI will help determine which roads are accessible during flooding and which can accommodate the evacuation traffic.

“These modules can be linked to existing real-time rainfall gauges and weather forecasts for improved accuracy and usability,” Long says. “The transportation safety or disaster planner can use these results to produce planning documents based on geospatial data and information to develop region-specific tools and methods.”

Long and Corns received $124,000 in funding from MoDOT and MATC for the 12-month project, which is in its early stages.

The project combines Corns’ expertise in computational intelligence and complex systems with Long’s focus on disaster recovery. In 2013, Long and Corns worked with the USGS and the University of Puerto Rico at Mayaguez to develop a model to help city planners return their communities to their pre-event state after a large-scale disaster. That project was inspired by the EF-5 tornado that ripped through the southwest Missouri city of Joplin in 2011.

Posted by

On August 2, 2019. Posted in 2019, Around the Puck, Research, Summer 2019