Software devises best plan for tackling forest fires

 作者:禄浴儒     |      日期:2019-03-02 05:07:00
By Tom Simonite Software that generates a plan of action for tackling forest blazes is being tested by fire-fighters in Spain. The system rapidly decides how best to allocate the resources needed to put out a spreading fire. “The idea is to support people fighting the forest fires that happen in the summer months,” says Luis Castillo, an artificial intelligence researcher at the University of Granada, Spain. He designed the planning system – known as SIADEX – with colleagues Juan Fdez-Olivares, Oscar Garcia-Perez and Francisco Palao. Computerised maps are already used by people in charge of managing the fire-fighting effort. These maps are used to plan which areas to focus on and which resources to deploy, such as fire engines, planes and helicopters. But working out the details of such a plan involves coordinating thousands of people, hundreds of vehicles and many other resources. SIADEX is able to help by rapidly weighing up different variables. For example, it calculates which fire engines could reach an area first, where aircraft could be used, and even how to organise the shift patterns of individual fire fighters. It then very quickly produces several different detailed plans. “It suggests two or three optimisations for approval based on different measures of success,” says Castillo. One plan might be the cheapest, another the fastest, and a third the least complicated. To devise a plan, SIADEX uses a technique called Hierarchical Task Network planning to breakdown the standard operating procedures already used by fire fighters into key constituents. More complex software is then applied to assess how these different parts might apply to the current situation. The final plan is translated back into terms that fire fighters can easily understand. Once a plan has been approved, everyone involved – from individual fire fighters to administrators – is informed of their duties via a website that can be accessed using a PDA, laptop or desktop computer. Emergency workers can also use this site to keep SIADEX informed about developments on the ground, enabling the system to modify its plans. “Users are prompted for feedback on what they see or do,” Castillo says. “For example, they are asked to click when a helicopter arrives.” In the summer of 2006, the researchers tested the information-gathering part of SIADEX during more than 100 fires in the Andalusia region of Spain. The full system has been tested in the lab and is scheduled to undergo outdoor testing during the summer of 2007. Austin Tate, who works on emergency planning systems at Edinburgh University in Scotland, UK, says a key strength of SIADEX is that it is “ready to go now, not in 20 years.” Another strength, he believes, is that it can be used without much training. “It uses a well-established planning technique that can be understood by the user,” Tate told New Scientist. Governments, emergency planners and armed forces around the world are waking up to the potential of systems like SIADEX, says Tate. “Homeland security, tsunamis and other emergencies need complex coordination and collaboration,” he explains. “People are desperate to get AI into these areas.” A paper describing SIADEX was presented at The International Conference on Automated Planning & Scheduling in Ambleside, UK, in June 2006,