The future of the world with artificial intelligence (195);
Engineers have made it easier to detect real-time flooding on roads by developing a machine learning model.
Road-related incidents are one of the leading causes of flood fatalities in the United States, but the lack of adequate flood reporting tools makes assessing road conditions challenging.
Existing tools such as traffic cameras, water level sensors and social media data can provide flood observations. However, they are often not primarily designed to measure flood conditions on roads and do not work in conjunction with each other.
A network of sensors can increase awareness of flood levels, but they are expensive to operate at scale.
Rice University engineers have developed a potential solution to this problem: an artificial intelligence-based automated data fusion framework called OpenSafe Fusion.
Open source situational awareness
OpenSafeFusion uses existing individual reporting mechanisms and public data sources to quickly understand road conditions during increasingly frequent urban flood events.
Jimmy Padgett, Stanley C. Moore, professor of engineering and chair of the Department of Civil and Environmental Engineering, along with Pranavesh Panakal, a postdoctoral researcher in civil and environmental engineering, analyzed data from nine sources in Houston before developing a comprehensive framework for an automated data system in their research study, “Eyes More on the Road: Sensing Flooded Roads Using Real-Time Observations from Public Data Sources”, published in the Journal of Safety Engineering.
“Resources that directly observe flooded roads are limited, urban centers are full of sources that directly or indirectly observe flood or road conditions,” Padgett said.
Padgett and Pannakal hypothesized that an automated system incorporating insights from these real-time sources could revolutionize flood situational awareness without significant investment in new sensors.
Padgett said the study provides a pathway for communities to respond to urban stressors such as flooding using existing data sources.
The system was inspired by our long-standing collaboration with our colleagues at the SSPEED Center in Rice, who are developing advanced flood warning systems. Here we focus on the impacts of flooding on transport infrastructure and on understanding how other data sources can complement the information from flood models with regard to the impact on roads and safe mobility.
Machine learning and data fusion
The framework uses data from sources such as traffic alerts, cameras, and even traffic speeds, and uses machine learning and data fusion to predict whether a road will flood.
The value of such data sources was evident during Hurricane Harvey in 2017. Many people in Houston—including emergency responders—have resorted to manually checking data sources to infer probable road conditions to overcome the lack of reliable real-time road condition data.
To test the OpenSafeFusion process, researchers used historical flood data observed during Harvey to reconstruct the scenario in this framework, which included approximately 62,000 roads in the Houston area.
Panakal said: This model was able to observe about 37 thousand road links, which is about 60% of the network we considered, and it is considered a significant improvement.
Other data sources that can be used in this framework include water level sensors, citizen portals, crowdsourcing, social media, flood models, and what the study calls humans in the loop.
Responsive artificial intelligence
Panakal says: This last source is very important; Because the human element of “OpenSafeFusion” enables the responsible use of artificial intelligence.
“We don't want a fully automated system without any human control,” Panakkal said. Therefore, we designed protections based on the responsible use of artificial intelligence. This need for responsible AI in such tools is still an open area for further work, and we hope to delve deeper by testing our methods in the future.
The study also considered the effects of flooding on community access to critical facilities such as hospitals and dialysis centers during natural disasters. This helps community members or emergency responders know which roads are flooded and how to get to a location safely, Panakal said.
Padgett says the researchers hope to pursue extensive testing, validation and exploration of how communities of different scales and access to resources can use the framework to better understand road conditions during floods.
“Due to the effects of climate change and intensified weather events, the frequency and severity of flood events could increase in the future, so we need a solution to better respond to flood events and their effects on infrastructure,” Padgett said.
