An ontology model for a context-aware preventive assistance system: reducing exposition of individuals with traumatic brain injury to dangerous situations during meal preparation

Severe traumatic brain injuries often decrease individuals' judgment and awareness. Ensuing cognitive impairments could then expose these individuals to dangerous situations. This paper presents a Preventive Assistance System (PAS) for meal preparation. PAS was designed to prevent a detected potential risk situation from progressing towards a situation that is dangerous for the user's safety. Its final aim is to avoid a drastic safety intervention by the autonomous Safety System. PAS provides preventive assistance to four out of the six potential risk situations identified by an interdisciplinary team. It relies on a model of assistance, communication, and smart homes. This model is described in a taxonomy implemented in OWL 2. The preventive assistance model is inspired from Hierarchical Task Network (HTN) and the communication model is inspired by the speech act theory. While observing the user, the generation of a preventive assistance tree is guided by the Instrumental Activities of Daily Living (IADL) being executed and the detection of potential risk situations. The preventive prompts -jingles, pictograms, voice messages, written messages...- are provided gradually, proceeding from implicit/generic to explicit/specific, until the user successfully reacts to avert the danger.