Abstract: |
This paper describes the application of object oriented genetic programming to the automatic generation of agents under the Object Oriented Paradigm. To generate the agent programs code, we evolve concurrently the methods that represent the agent-environment interaction. We use like terminals and operations the objects that correspond to the context elements. This study uses the simulation league of the Robot World Cup (Robocup) like a testing environment. The fitness function used evaluates the behavior of agent player in several levels that indicates the learning progress. The experimental results indicate that is possible the agent programs evolution under the Object Oriented Paradigm. |
