A Context-Driven Complex Activity Framework for Smart Home

Internet of Things (IoT) will provide a data rich world to afford smart systems which, in a context of a smart home, will have to adapt to people according to their needs for work and living. This paper describes the work to establish a framework for building a database to capture all possible user interactions associated with a given activity and lists the conditions for these user interactions leading to the activity goal. As an application of this framework, an activity performance-based supervised recommender system to recommend tasks related to different activities in a smart home is presented. Its usefulness to enhance the quality of life of people in technology-laden surroundings can be seen, especially for the elderly people, by providing environments with smart agents, that can effectively assist them in user interactions, as well as recommend tasks or activities, or even guide them through the intended tasks. Furthermore, the future intelligent assistant agents will have to learn a universal set of user interactions related to any activity to effectively mitigate distractions and adapt to the situations and recommend tasks to aid users in reaching to the end goal of performing the desired activity. A preliminary result demonstrates a performance accuracy of 75% when evaluated on a subset of the UK DALE dataset [1].