Teaching drone swarms to assist in search and rescue efforts
Category: research Video duration: Teaching drone swarms to assist in search and rescue efforts
Ryan Williams, assistant professor of electrical and computer engineering, is working to teach multi-agent systems, or drone swarms, how to work with search and rescue personnel to improve real-life, on-the-ground searches in Southwest Virginia.
Thousands of people go lost every year in wilderness environments. Around 25 percent are either never found or when they are found her hurt. So if someone that is lost in the wilderness, we would like to not only deploy humans as we already do, but we would like to deploy aerial vehicles in concert with human teams. And this is not about replacing human teams with aerial robots. It's about complementing human teams because we have trained experts that are really great at finding those people and extracting them. So you cannot replace that capability. But with an aerial robot, we can cover a lot more grounds. And we have different sensors that can supplement a search. So that's at a very high level. What we're looking at is the point autonomous aerial vehicles to support human teams. And there's a lot that goes into that when it comes to the economy, That's really the key here. Because we're not looking at remotely piloted aerial vehicles. What we would like to do is have the, the aerial vehicles make decisions for themselves. So we equip all of them with high-quality thermal cameras for two reasons. Number 1, we can see hot bodies. If I wanted to differentiate between say a person, any tree, a thermal camera will help tremendously with that. But more importantly, we're tackling a roof wood is really a tough problem, is in wilderness environments you have to deal with the trees, you have to deal with canopy. Trying to see through the canopy completely is a bit tough, but with thermal with sparse canopy, you can actually see somewhat through the canopy and see those hot bodies. They do cooperate. So that's another big problem that we're solving is the information problem. So with more vehicles, you gather more information and not only give the process that information, but you have to move that information between vehicles. Because if the vehicles are separated by large spatial scales, it helps for them to know what the other vehicles are doing. That's why this search and rescue problem really turned into great domain for a lot of really tough research problems that are cross discipline. So with communication types of problems, we have information processing types of problems. We have computing types of problems, we have autonomy problems. We have modeling problem. So we're, we're creating new search models, new loss person model. So we can try to predict what a lost person than they do in a wilderness environment. So it's really intersecting a lot of tough and interesting research problems. I think we're both a mixture for a very, very good search and rescue solution.