Problem Statement

Most smart home systems rely on fixed schedules and commands set by the user. This means users have to manually decide when lights, fans, ACs, and other appliances should turn on or off. However, people do not follow their routines at exactly the same time every day. For example, someone may turn on their bedroom light around 8 AM, but the exact time could vary from day to day. Existing automation may not adapt well to these small changes, making the system less convenient and requiring repeated manual adjustments. This can also lead to appliances being left on unnecessarily, resulting in avoidable energy use. The problem is that smart homes can follow instructions, but they often do not learn from the user's everyday behaviour. There is a need for a system that can understand repeated usage patterns, suggest flexible automations, and still allow the user to decide whether those suggestions should be applied.

Solution

AI-powered home automation that learns user habits and suggests flexible, personalized appliance schedules.

In Detail

Our solution is an Adaptive Home Automation system that learns the user’s everyday appliance habits instead of depending only on fixed schedules. Whenever a user operates an appliance, the system records details such as the appliance, ON/OFF status, date, and time in MongoDB. After collecting enough data, it identifies repeated patterns and estimates a flexible time band rather than assuming one exact time. For example, if a bedroom light is usually switched on between 7:50 and 8:10 AM, the system can recognize this as a routine and suggest automating it. Gemini Multimodel API helps turn these patterns into simple, personalized suggestions. The user can accept or reject each suggestion, keeping them in control. Once approved, an ESP32 connected through a web server can control appliances such as lights and fans. This can also help reduce unnecessary energy usage by adapting automation to real behaviour.