Your Rescuer in the Future May Be ... a Robot?
CREDIT: National Science Foundation
In the future, if ever you find yourself in a house on fire, perhaps a few of your emergency responders will not be people at all.
Rather, they could be expertly programmed circuit boards with a knack for analyzing events around them without any help from humans and cooperating with one another to respond in the most effective manner. In other words, they may be robots that are independent, adaptive and cooperative.
This is what two University of Delaware professors mechanical engineer Bert Tanner and linguist Jeffrey Heinz are working on with support from the National Science Foundation.
"If you have an idea about how a fire is likely to spread, you reposition your resources to be as effective and efficient as possible," Tanner told InnovationNewsDaily. "Experienced humans can do that almost subconsciously. If we can enable machines to do that, it can lift the pressure off the shoulders of human supervisors."
Analyze and respond
There are two key abilities the robots must possess to make this possible, Tanner said. First, they have to learn about their world and adapt their behavior to it. And second, they have to coordinate with one another.
To achieve the first goal, the robots need to continuously absorb and use information from their environment to gain a better understanding of that environment.
"The idea is that a robot can build an abstract, not very detailed model of a particular phenomenon in the environment and incrementally improve the model based on observation," Tanner said. "The challenge is to be able to build a model of what is happening around you quickly, as you go, refine it as more information becomes available, and use it to compute an even better plan of action."
In the real world this might mean predicting the future path of a fire based on the pattern of how it has been spreading so far. Next, the robots need to coordinate with one another and coordinate their abilities in an efficient plan that will achieve their mutual goal (just like Tanner is using his engineering skills and Heinz his linguistic knowledge to design cooperative, adaptive robots ).
"Robots should be able to 'understand' how their actions are interdependent, and how the activity of one may enable or inhibit the activity of another," Tanner said. "We're working on the mathematical tools to express accurately how distinct robots' behaviors interact."
Borrowing from language
To enable the robots to coordinate effectively, the researchers are using what they know about the structure of human language to develop the language of these robots.
We all know that every language has its own set of rules about when sounds or letters may or may not be combined in English, for example, "ng" is an acceptable way to combine two letters, but you'd never find them at the beginning of a word. Similarly, a robot "knows" in what order it is allowed to perform certain actions. For example, a robot with the ability to move , grab and release something, "knows" it can perform these three actions only in a particular sequence to achieve a desired outcome.
"The way that certain sounds go together and certain others don't in the words of a natural language seems to have a lot in common with how certain actions by one machine can enable or inhibit other actions by the same or another machine," Tanner wrote.
Once the researchers have completed their design, Tanner hopes it will allow policy makers to develop more efficient plans for specialized teams of humans, and hopefully, machines, to respond effectively to emergency situations .