Abstract: With the intensification of global population aging, the demand for elderly healthcare is increasingly urgent. Traditional health monitoring systems rely on wearable devices or single-point sensors, facing issues such as low compliance, data isolation, and poor user experience. Proposes an AI-driven multi-agent healthcare system in smart home environments, utilizing smart appliances (e.g., refrigerators, mattresses, toilets) as behavior-sensing nodes to collect real-time behavioral data (e.g., usage frequency, pressure distribution, toileting patterns). The system employs the LangGraph framework for multi-agent task coordination, integrating modules for health monitoring, device control, and behavior understanding, while analyzing health trends through expert rules and machine learning. By leveraging the “appliance-as-sensor” concept and user-friendly notification mechanisms, this system significantly enhances elderly user experience and addresses accessibility and privacy concerns, offering a novel solution for integrating smart homes with elderly care.

Key words: Smart home, AI multi-agent, Elderly health, Smart appliances, Health data analysis, LangGraph framework, Behavior recognition, Health assessment