There’s been a lot of interesting talk about “homo-economicus” on the blog this semester, so I figured it would be fun to close things out with a post on “robo-economicus.” That is, robots using strategies modeled on economics to cooperate. Let me explain…
A few months ago, I was applying for REU positions in mathematics and computer science. One of the programs I considered was Oregon State University’s Robotics in the Real World. In doing my due diligence in preparing an application, I stumbled upon this paper, “Learning to trick cost-based planners into cooperative behaviour.” In the paper, the authors essentially share a drone traffic-control system in which a central algorithm determines a pricing scheme for various pockets (“sectors”) of airspace that the drones pass through. Drones are then left to make their own decisions based on that pricing scheme. They report pretty good performance of a model of the system.
Multi-robot coordination is a tricky problem. Often, it becomes computationally infeasible to centrally plan the actions of each agent, especially when working with large number of robots. Instead, control must be distributed among the robots. But, engineers still want the m working cooperatively, not just all doing their own thing. The market system seems to have provided a pretty good solution to this problem for people, so many researchers are working to employ elements of that system in robotics problems. A quick google search for “market economics multi-robot coordination” will turn up much more material for your perusal, if you’re interested.