Retrofit Conventional Appliances Control Using Machine Learning Techniques

Abstract

Internet of Things (IoT) concept has been widely used in various devices such as domestic appliances, energy meters and lighting equipment for remote control and monitoring. Smart household devices, such as the smart thermostat, smart speaker or smart refrigerator, have aroused consumers’ interest owing to its integration of Machine Learning (ML) leading to a higher degree of home automation. It is based on the fact that the usage patterns are collected, analyzed and learnt during the operation. However, the user who wishes to enjoy the new features associated with the ML has to dispose of the old, non-internet-enabled appliance, even though the old machine is still well functioning. In lieu of creating undesirable wastage, this project has an aim of retrofitting existing domestic appliances to become Internet-enabled devices, with the integration of ML. The conventional air-conditioner is chosen as an example that adopts ML. The fundamental concept of this project is to deploy several sensors to collect temperature, relative humidity and indoor air quality readings. ML-based computer vision is also used to count the number of people who occupy in an indoor zone. Three air-conditioners are controlled simultaneously according to the user temperature setpoint. A prototype device involving sensors, IR Transceiver and Human Machine Interface has been built. The result shows that the proposed system regulates the temperature in higher precision over the conventional method. The user can evaluate the system status using a web-based dashboard.