David Shaffer: Emergent behaviors in adaptive agents

I have been studying emergent behavior in simple robotics systems (simulated). I am interested in robust training of teams of simple robots for the completion of tasks which require cooperation.

Student projects

Amanda Gentzel (2011) studied using a modified version of the genetic algorithm to train simulated robots to complete a "box pushing task." Amanda's simulation was spatially discrete and did not yield very interesting emergent behaviors but there were some interesting results which might help future projects. Read more about this project on Amanda's honors project page.

Brad Patton (2005) studied using the traditional genetic algorithm to train teams of robots to play freeze tag. Brad's project details are on his wiki page

Bryan Mackrell (2002) developed a 2-d robotic simulation framework in Smalltalk. In his capstone project he further developed this framework to include web-based competition and to use a distributed objects (CORBA) for robotic control. This gave the ability for people to create robotic control algorithms and test them in a remotely simulated environment, possibly against other remotely-controlled robots. Bryan's work also served as the launching point for Brad Patton's honors project (above). Bryan's wiki page has lost its images but still might provide some information to the interested reader.