Internet of Things-Driven Smart Furniture Systems for Human-Centered Personalization: Experimental Evaluation Using...

Area of Science:

• Human-Computer Interaction
• Artificial Intelligence
• Internet of Things

Background:

• Traditional furniture lacks adaptive capabilities for individual user needs.
• Existing smart furniture systems often have privacy concerns and limited personalization.
• Optimizing ergonomics and energy efficiency in furniture requires advanced control strategies.

Purpose of the Study:

• To develop an advanced Internet of Things (IoT)-driven smart furniture system for dynamic user adaptation.
• To integrate deep reinforcement learning and federated meta-learning for personalized furniture experiences.
• To enhance human-furniture interaction in health-aware environments.

Main Methods:

• Formulated personalization as a Markov decision process for sequential adjustments.
• Applied an adaptive Kalman filter for real-time estimation of ergonomic preferences.
• Utilized a sparse autoencoder for 82% reduction in sensor signal data while preserving temporal features.
• Implemented federated learning (FL) for privacy-preserving collaborative training across distributed units.

Main Results:

• Reduced cumulative user dissatisfaction by 43% and energy consumption by 21% compared to rule-based systems.
• Achieved real-time adaptations with an average latency of 280 ms, upholding ergonomic constraints in 95% of use cases.
• Federated learning training converged to 87% of global performance within 30 iterations without raw data exchange.

Conclusions:

• The proposed framework offers a robust, responsive, and interpretable solution for smart furniture.
• Demonstrated significant improvements in user satisfaction, energy efficiency, and ergonomic support.
• The system's privacy-preserving and scalable nature makes it suitable for health-aware workspaces, smart homes, and eldercare.