People construct routines as they repeatedly perform the same sequence of actions. Routines provide a huge benefit by freeing people?s attention, allowing them to carry out their daily tasks without constantly thinking about every little thing they must do. Problems begin to arise when people must deviate from their routines. Families rely heavily on their routines to address the complex logistics and conflicting agendas of work, school, family, and enrichment activities. However, families often deviate from their routines, and when breakdowns in the plans occur, they feel their lives are out of control.
This research will develop a system that learns the routine movements of family members, and a planning system that leverages this model in order to generate a speculative plan for future days. The system will also predict conflicts with scheduled deviations and detect when plans begin to breakdown, such as when someone forgets to deviate from a routine. A calendar interface that displays the routine movements of family members along with their scheduled deviations and a small set of reminder applications that help people enact their plans and that support them when plans breakdown will form the basis for evaluating the underlying systems. This research is transformative in the novel integration of machine learning and planning techniques, and its application to a real-world and complex problem. Finally, this research provides insights on how intelligent, ubiquitous computing technology influences families? feelings of control and their quality of life.
The proposed work has the potential to significantly improve the quality of life for millions of families by reducing stress caused from breakdowns in plans and routines. Lowering stress can improve the quality of marriages, the quality of parenting, and the physical and mental health of children. We will involve undergraduate and graduate students in our research and will incorporate our findings into our courses on ubiquitous computing, interaction design, and on smart homes. We expect that our focus on a social problem will attract non-science-focused students to science and expose science-focused students to design methods of inquiry.
This project has two major goals. First, we want to understand what kinds of routines people engage in that would benefit from computational assistance. Second, we want to build systems that can extract routines from individual and group sensed behaviors, and then use knowledge of these routines to support future planning tasks. We have demonstrated that we can extract routines that provide value to people, despite the fact that routines are idiosyncratic. In particular, we demonstrated that we could extract routines about how parents coordinate to pick up and drop off their children, with the goal of reducing stress. We have also shown that we can use routines to predict when a person will return home to more efficiently and effectively control heating and air conditioning systems for reducting energy usage, and to predict movement through a home to control heating and air conditioning systems within zones of a home. We have built tools that allow us to extract these routines across a variety of behaviors and domains. This includes algorithms and visualization techniques that will support scientists in a variety of domains to identify routines and build interventions that can help individuals and groups. This work has supported the thesis work of multiple PhD students and has supported a number of undergraduate and Masters students.
