The Air Force wants to enlist robots to fight forest fires.
A new competition aims to test the ability of artificial intelligence — generated by teams in the United States and the United Kingdom — to plot search-and-rescue missions, among other tasks. The competition, which opened Jan. 15, and is expected to conclude March 31, will reward the teams with cash, travel and possibly a trip to Washington, D.C., and meetings with the Air Force. For the military, it might gain another new tool for managing swarms.
The competitions are put on by the Air Force Research Laboratory, United Kingdom Defence Science and Technology Laboratory, or DSTL, in collaboration with Wright Brothers Institute and University of Dayton Research Institute. Using videos provided by the United Kingdom Forestry Service, the teams competing in the parallel competitions on both sides of the pond will program AI for simulated swarms of robots. The simulated drones will all be based on an AFRL platform and use the same sensors so that, like in stock car races, the competition focuses entirely on the differences in how the vehicle in controlled and not disparities in how the vehicle is built.
While the overall competition is billed as primarily about search and rescue in a natural disaster situation, both the Air Force and Wright Brothers Institute announcements note that these technologies may have defense applications.
That’s an understated way to put it.
Wildfires are, perhaps, the natural disaster that best maps onto a dynamic battlefield, especially one created suddenly by an insurgency. Using drone swarms to autonomously map the points of interest, changes in the presence of danger and where civilians are could help in both kinds of firefights. Anticipating the spread of fires is aided by having information delivered in as close to real time as possible. Fires may appear suddenly and move at great speed, but the movements are dependent on where the fires start and the conditions of the environment in which it is trying to spread. And, of course, forest fires and battlefields are hardly exclusive events, so knowing how to handle the former still has relevance for commanders in the field.
Once in the competition, the teams will go through three phases. The first will feature the easier challenges and teams will be allowed to upload their results as many times as they like, while being judged only on the highest submitted score. That phase will run until Feb. 15. The second phase will see a harder, more complex challenge, which will run through March 15. After March 15, the field will be winnowed to the top 10 teams, who will be invited to compete in the final showdown, where their drone-code will be scored in real time.
Dynamic swarm operations are a new enough field that competitions like this seem as good an opportunity as any to shape the future of robots finding people in war. Or in fires. As the Pentagon flies an increasing number of drone missions over the United States in support of firefighting efforts, it wouldn’t hurt for those drones to know what they’re supposed to be doing.