Modernizing Smart Homes into Cognitive Residences

The growing popularity of connected homes can be attributed to advances in cloud computing and machine learning. The latest technology allows devices such as smart speakers and thermostats to learn from your feedback and adapt to your personal preferences. However, these smart gadgets can also influence people's behavior. Investigate the potential impact of smart homes on behavior using a human model with hierarchical reinforcement learning. This model can be used for a variety of tasks, such as controlling humidity and temperature within a smart home under Q-learning management. We investigate possible behavioural changes and explore prospects for mutual adaptation between smart home and human models. Our human models are trained on these tasks using hierarchical reinforcement learning, which allows them to learn when to move on to new activities and how long to focus on existing activities. We then integrate human modelling with digital house models in real-world scenarios and conduct in-depth studies involving different situations in which one or more human models interact with a smart home. Our results suggest that installing a connected home may cause human models to exhibit unexpected behavior, such as frequent changes in activity or longer times to change temperature settings. What is interesting is that when two human samples are combined, some exhibit typical behavior while others reproduce the surprising results of single-subject experiments.