This paper presents a novel recommendation methodology for learners who are doing tasks in ubiquitous learning environments. The main concept is to deal with a learning task as a set of real objects and to orient the recommendation methodology towards these objects. These objects can be detected using ubiquitous technologies. An intelligent agent can propose diverse recommendations for the learners based on the objects' attributes. These recommendations include assistance from peers, educational materials and learning tasks. In order to do this, the agent constructs and maintains a learner model based on the learner's experiences in using the real objects that conform a task. The proposed ubiquitous learning environment model, for learner modeling and recommendations, is a belief system implemented in DLV under the ASP (Answer Sets Programming) paradigm.