Population ageing is widespread across the world. Unprecedented in the history of mankind, this demographic trend leads to a number of social and economic issues related to ageing/disabled people whose number increases considerably over the years. As the number of caregivers can not evolve accordingly, we must now think of alternatives allowing these people to stay longer in their preferred domestic environment with a sufficient level of autonomy. Home automation, which includes computer, electronics and telecommunications within a housing, allows to provide technological solutions. In this work, we propose a model-driven flow for assistive home automation systems. First, we seek to integrate a home automation non-expert (e.g. familiy member / caregiver, occupational therapist) in the design process to better take into account the needs of people with disabilities, but also to limit the costly intervention of the expert on site. To do so, we propose a modeling language (DSML: Domain Specific Modeling Language) allowing one to describe a home automation system by manipulating high-level abstraction concepts. This DSML is composed of two views : the environement view to model the home automation environment and the interaction view to model user-system interactions. In this latter, two types of interactions are defined to consider the greatest number of needs. Thus, it is possible to define requests (i.e. services) and temporary prohibitions (i.e. modes) on home automation devices. From models defined with this language, our design flow allows to generate the control part through successive model transformations. An evaluation on the Kerpape Smart Home case study with students and occupational therapists provides interesting results on both the proposed design flow and the DSML usability. Then, we seek to offer an alternative to context-aware approaches based on the massive deployment of sensors, which can sometimes be problematic when one focuses on the user acceptability (cost, intrusivity). We propose to integrate in the design flow a context-aware model based on ontology to define user-system interactions. Unlike existing solutions, this model is not based on sensor data, but on the semantic analysis of each user request on home automation devices. The ontology and associated rules allow an inference engine to extract information usually sensored: according to the way the user interacts with his home automation system, his location and his intention can be deduced by logic reasoning, leading to contextual services proposition. An evaluation on the Domus Smart Home dataset shows the interest of such an approach to limit the use of sensors, but also to increase the dependability by information redondancy. Part of this work is integrated in the ASIM project supported by Vity Technology. The concrete syntax (i.e. graphical representation) of the modeling language will especially be considered in the purpose of a commercialization.
