Towards a Simulation Framework for Smart Indoor Spaces

The effectiveness of sensor-based applications for smart homes and smart buildings is conditioned upon the deployment configuration of their underlying sensors. Real-world evaluation of alternative possible sensor-deployment configurations is labor-intensive, costly, and time-consuming, which implies the need for a simulation-based methodology. In this work, we report on such a methodology that supports the modeling of indoor spaces, the activities of their occupants, and the behaviors of different types of sensors. We argue that, in order for a simulation to be useful for the purpose of evaluating a sensor deployment configuration, it has to generate realistic event streams of individual sensors over time, as well as realistic compositions of sensor events within a time window. We have evaluated our simulator for smart indoor spaces, $SI{M}_{sis}$ toolkit, in the context of our Smart-Condo ambient-assisted living platform, supporting the observation and analysis of activities of daily living (ADLs). Our findings indicate that $SI{M}_{sis}$ produces realistic agent traces and sensor readings, and has the potential to support the process of developing and deploying sensor-based applications.