Subsumption architecture for motion learning in robots

Intelligent control architectures use computational intelligent techniques in order to improve robots performance. This paper shows a control architecture based on subsumption, that uses some computational intelligence techniques. This architecture provides to a robot the ability to learn how to perform a set of specific motions. A genetic algorithm is used to find the adequate robot movements, and then a set of neural networks are trained to learn those movements. We conducted a set of experiments in a robot simulated environment, in order to show the performance of the control architecture in every one of its stages. Results show that the proposed architecture is able to learn and perform basic movements of a robot independently of the environment or the robot defined structure.