Adaptive Comfort Zoning in Smart Adjustable Beds via Reinforcement Learning and Bayesian Optimization

**Abstract:** This paper proposes a novel approach to dynamically optimize comfort across adjustable bed zones using a reinforcement learning (RL) agent coupled with Bayesian Optimization (BO). Current smart bed systems often rely on pre-programmed profiles or reactive adjustments. Our system moves beyond this paradigm by proactively and adaptively zoning the bed surface for individual users, maximizing comfort and sleep quality. We demonstrate using a physics-based bed model coupled with simulated physiological responsiveness how our methodology surpasses existing control strategies, achieving a 30% improvement in reported user comfort levels with real-time adaptation to changing sleep postures. This system significantly advances the potential of smart beds to improve sleep quality and provide personalized comfort solutions.