One important research area in autonomous mobile robotics is to create companions that live in our ambience and perform tasks to help in everyday life. On the other hand, Ambient Intelligence (AmI) seeks to create a network of helpful intelligent devices. This paper describes an approach for the development of cooperation models for eldercare robot teams using goal-driven control components. The framework and the approach are illustrated through the development and assessment of task allocation in multi-robot teams. Two cooperation models are implemented: (i) a team model based on the adaptive multi-agent systems theory where task responsibility is agreed among team peers by exchanging individual estimations of the degree of difficulty and priority to achieve the task; (ii) a hierarchical model where a robot manager asks for the estimations of its team members and then assigns the task. Experimentation for team cooperation assessment is performed through considering environmental changes, as well as communications and internal failures. The proposal is simulated in an AmI-oriented elderly care center to assist seniors in need.