Current intelligent landscape environmental comfort adjustment is not sensitive to the physiological state of users and dynamic needs, and cannot achieve adjustment in time and personalization. This paper proposes a dynamic adjustment system of intelligent landscape environmental comfort designed on human-computer interaction. The system synchronously gathers parameters of the environment (temperature and humidity, lighting, noise) and user physiological behavior (heart rate variability, skin conductance, gait) by means of a heterogeneous sensor network. It uses the LSTM random forest hybrid model to implement multi-source data fusion and comfort prediction, and uses reinforcement learning algorithms to construct optimization control instructions of landscape equipment (shading, irrigation, lighting), which form a "perception analysis-decision execution" closed loop. The system has achieved 7.6 points average user comfort rating in 7 typical scenarios, 91.8% average personalized demand satisfaction rate, and 9.6 kWh average daily energy consumption, which has verified the effect of achieving accurate comfort improvement and energy efficiency synergy through dynamic interaction.
Zhao et al. (2026) studied this question.