fbpx

Artificial Intelligence, News, Technology

Using AI for good: Algorithms for disaster planning and response

Amie Haven

Amie Haven, Journalist
@uxconnections

Researchers at the USC are using AI to preserve water supplies after quakes and help the US Coast Guard prepare for disaster relief.

Researchers from the University of Southern California (USC) are working on artificial intelligence (AI) systems to assist in disaster planning and response. Within the USC’s Center for Artifical Intelligence in Society (CAIS), principle investigator Bistra Dilkina and co-investigator Milind Tambe are using AI to make disaster planning and response more efficient. 

Planning for and responding to disasters is a complex effort, with many teams needing to coordinate to provide the best knowledge, skills, and resources in a timely manner. Efficient preparation and execution in times of crisis means more lives saved. Climate change has meant that the world is seeing an increasing number of natural disasters which can uproot communities and devastate lives. The researchers say that designing a disaster planning and response system which mines large amounts of complex data – locating the most salient information – means resources will be allocated more efficiently. The objective is to save both lives and money.

Dilkina says, “Leveraging AI for disaster planning and relief is a critical component of computational sustainability—the development of models and algorithms that address high-impact social and environmental problems”. Previous work in Senegal, West Africa led to the creation of an AI framework to support flood-resistant road infrastructure planning. The AI framework combined flood prediction models, algorithms to determine optimal resource allocation, and data from phone calls made during previous floods to determine mobility needs. Overall, the AI framework delivers information to form a complete picture. With limited budgets in mind, the information can help determine which stretches of road are most in need of fortifying to ensure flood resistance whilst increasing mobility for communities. 

Current projects in Los Angeles (LA) in the US involve securing water pipes for resilience in future earthquakes. The US Geological Survey (USGS) and the Southern California Earthquake Center (SCEC) estimate that within 30 years there’s a 1 in 5 chance of a 7.0 quake along the San Andreas Fault beneath California. Massive quakes can disrupt essential water supplies, leaving homes and hospitals without water whilst firefighters are left unable to tackle blazes.

The USC’s project uses data and maps of LA’s water system to create computations and scenarios which identify the most efficient areas for quake-resistant pipes. Replacing the entire system isn’t financially viable so it’s crucial to target specific areas, ensuring that essential supplies of water are preserved during a disaster.

This form of organising complexity is also benefiting the US Coast Guard. Dilkina and Tambe are developing algorithms that will help determine the most efficient allocation of manpower and equipment during crises. In times of competing demands, disaster relief teams need to know that they are providing what’s needed whilst keeping some resources back in case another disaster strikes. In the devastating earthquake in Haiti in 2010, where around 250,000 people died, the US Coast Guard were some of the first responders. However, those first teams were unprepared for the situation and found themselves lacking the equipment and manpower to fully help those in need. An AI system which can rapidly calculate need by sorting through huge amounts of data could be a real lifesaver in future scenarios. 

Dilkina plans to scale-up the project so that is has global relevance and to ensure equitable use. The team plan to use AI for social good and Dilkina says they hope to “enable better disaster preparation and response as well as mitigate the impacts of natural and man-made catastrophes.”

Leave a Reply

Your email address will not be published. Required fields are marked *