This paper presents a novel anthropomorphic multi agent hybrid architecture with abilities of self-awareness and cognitive control for homecare robots. The paper presents the details of the robot memory (knowledge/data base), self-awareness model, learning method for weight updating, and the decision making algorithm. The developed architecture will facilitate a robot to adapt to an uncertain and dynamic home environment and improve its intelligence, reliability and flexibility for effectively performing homecare tasks.Index Terms-homecare robot, multi-agent architecture, self awareness model, robotic learning, robotic decision making.