Hotel room design affects the customer experience and hotel room design is highly dependent on experience of manual. It is inefficient and personalized poorly. This paper aims to present an intelligent hotel room design generation algorithm based on reinforcement learning (RL). The algorithm optimizes the spatial layout automatically by environmental modeling and intelligent decision-making. First, a guest room simulation environment is established. The design elements such as the furniture arrangement, traversable range and daylighting condition are quantified into the state space, and then design aesthetics and design function are defined as reward functions. Second, Deep Deterministic Policy Gradient (DDPG) algorithm in Deep Reinforcement Learning (DRL) framework optimizes the design solution by continuously simulation and feedback. When training the algorithm, the average design score increases greatly from 60.8 to 92.1, and the speed of convergence is 35.7% faster than traditional genetic algorithm. The experimental results show that the algorithm can generate the room layout satisfying comfort, design aesthetics and energy efficiency requirements within 3 minutes. The average user score is more than 86 points in N=120 survey.
Fu et al. (Thu,) studied this question.