A Smart Home Simulation Tool to Support the Recognition of Activities of Daily Living

Abstract

The prevalence of Internet of Things (IoT) technologies and internet-connected devices introduces the potential of providing intelligence to a wide range of applications. The smart home is a popular application of IoT. Smart homes are perceived as a promising solution for providing support to inhabitants in completing daily activities, prolonging independence and enhancing quality of life. Access to sensor datasets is essential for smart home research. For example, activity recognition research requires sensor datasets to assess the performance of activity recognition models and algorithms. There are several limitations that can prevent researchers from acquiring sensor datasets, such as cost, time, and inflexibility of smart homes. The limitations also extend to both the limited quality and quantity of existing sensor datasets. Smart home simulation is identified as a potential solution to mitigate these limitations. This paper proposes Smart Environment Simulation (SESim), a simulation tool that supports smart home simulation for generating synthetic sensor datasets. The paper discusses the design, implementation, and validation of SESim. SESim was validated by conducting three phases of validation. The first validation phase validated the utility of SESim by addressing several use-cases of smart home simulation. The second validation phase validated the utility of sensor datasets generated using SESim by conducting an experiment. The experiment involved the generation of synthetic sensor datasets using SESim, training of an activity recognition model using the generated sensor datasets, and the evaluation of the activity recognition model using the generated sensor datasets. The third validation phase assessed and validated the extensibility of the SESim tool by conducting an interview with an expert user who extended the functionality of SESim.