In this paper we provide our general approach and discuss relevant issues in providing a dynamic user modelling approach for attending functional diversity for accessible lifelong learning (ALL) in Higher Education. Our approach to provide universal and personalized access lies on combining user modelling and machine learning techniques to cope with the needs for ALL with a pervasive support of standards and supporting the full life cycle of service adaptation. The modelling differs from others in i) coping with interactions and context of the user that can only be considered at runtime and ii) characterising interaction capabilities of different kinds of devices. Models are used to personalize and adapt learning materials, pedagogical settings and interactions in the environment to satisfy both the individual learning needs and the access preferences, taking into account the context at hand.