Reinforcement Learning Based Particle Feedback Optimization in XR Heritage Interactions

With the widespread application of XR (extended reality) technology in the digital dissemination of intangible cultural heritage, improving user interaction immersion and feedback response quality has become a key issue in current research. Traditional interactive particle systems are mostly based on static rule settings, and it is difficult to achieve adaptive feedback adjustment according to user behavior, which limits the naturalness of the interactive experience and the integrity of cultural expression. This paper proposes a particle feedback optimization method based on deep reinforcement learning, constructs a policy network model that integrates user behavior state and feedback control parameters, and conducts systematic experimental evaluation in typical intangible cultural heritage XR scenarios. The experimental results show that this method is significantly superior to the traditional feedback mechanism in terms of immersion score, feedback naturalness, and system response delay, showing good stability and versatility. The study realizes the transformation of particle feedback from rule-driven to intelligent adjustment, improves the dynamic adaptability and cultural semantic expression level of the intangible cultural heritage interaction system, and provides a new path for the optimization design of immersive human-computer interaction systems.