User Portrait-Driven Smart Home Device Deployment Optimization and Spatial Interaction Design

This study proposes a deployment optimization model tailored to personalized smart home environments, guided by user profile data. The model integrates multi-dimensional inputs, including static attributes (n=15), behavioral sequences (length L=96, embedding dimension d=128), and preference tags (vector size k=64). It aims to enhance both device placement efficiency and spatial interaction effectiveness.The approach leverages dual-stream feature encoding, a reinforcement learning policy network, and a synergistic mechanism combining Gaussian Process Regression (GPR) and Q-learning. Experimental results indicate that the optimization reduces average communication latency from 61.2 ms to 38.5 ms and decreases energy consumption by 22.4%, demonstrating significant improvements in system responsiveness and energy efficiency.Based on the behavioral data of 500 users, the experimental results demonstrate that the proposed method significantly improves interaction accuracy (up to 93.1%) and resource scheduling efficiency (hit rate increased to 91.2%), compared to baseline strategies.