The autonomous robots are currently taking an important role in society. They can provide with help at home, hospitals, companies and industrial environments, as well as space missions in an autonomous and intelligent way through the implementation of cognitive architecture.This thesis comprehends the study from the design of cognitive systems for autonomous robots and proposes a cognitive architecture, compounded by agents, for the choice selection in autonomous robots with the purpose to excel the knowledge. This knowledge refers to all the criteria used to choose the most convenient task as the evaluation fro the mission objectives. This way, the robot could perceive and behave in the real world, learning from the relationship with its surroundings and using this experience in the process to planify, control and act depending on the most efficient choice selection.The architecture proposed comprehends, among others, one agent that implements an informatics ontology to represent all the necessary in a semantic process; a mechanism on the choice selection or a task manager from the robot during a certain period or to fulfil a mission and an objective function based on a "scale of values" that affect in the quality of the actions taken from the robot.The proposed model is flexible enough as it can build up with new knowledge, that will be classified with the default ones, category and affecting the decision making for e new value or specific attribute.The thesis develops different prototypes that allow validation the architecture proposed.