Artificial Intelligence for Smart Home Internet of Things

Abstract

Artificial Intelligence and Smart Homes have become increasingly prevalent in recent years. Smart Homes contain Internet of Things devices that generate large amounts of data, which can be used in Artificial Intelligence decision-making processes. This thesis researches how Artificial Intelligence is used in Smart Homes and the trade-offs between using Cloud Computing vs Edge Computing for AI. Two separate literature reviews were conducted to research these problems, screening the first 50 results from four databases for each review and analyzing the most relevant publications in depth. In total, 12 publications were analyzed in depth for the first research question and 8 for the second. Additionally, we developed our own machine learning application using Generative AI for an existing Smart Home environment. The application was developed almost entirely through Generative AI (Claude Code) with no manual changes to the source code. The application was deployed on a Raspberry Pi 5 with comparable performance to a Desktop PC. The trade-offs we found between Cloud and Edge Computing included trade-offs between computational power and latency, bandwidth usage, model training and inference, and privacy and security. However, few of the reviewed publications provided measured, empirical data to support these comparisons. The main use cases for Artificial Intelligence was found to be Smart Homes are energy management, elderly- and healthcare, home automation and anomaly detection. Overall, the results show that Artificial Intelligence has practical and varied applications in Smart Homes, and that both edge devices and Generative-AI-assisted development are mature enough to support them.