一种基于深度强化学习的VR重定向行走方法、系统
By constructing a deep reinforcement learning method with multi-scale observation space vectors and a multi-objective joint reward function, the problem of co-optimizing the continuity of exploration and the imperceptibility of perception in VR redirected walking is solved, enabling users to explore safely and continuously in the virtual environment and improving VR immersion and user experience.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SHANDONG UNIV OF FINANCE & ECONOMICS
- Filing Date
- 2026-04-24
- Publication Date
- 2026-07-17
AI Technical Summary
Existing VR redirection walking methods based on deep reinforcement learning cannot achieve the synergistic optimization of exploration continuity and perceptual imperceptibility when the virtual space is much larger than the physical space. They also ignore the differences between different features, which prevents the model from making full use of differentiated features for effective guidance. Furthermore, when focusing on obstacle avoidance tasks, they neglect the critical balance between walking safety and perceptual imperceptibility during redirection.
By performing dense and sparse ray sampling on obstacles inside and outside the target VR user's field of vision, a multi-scale observation space vector with asymmetric representation is constructed. This vector is then encoded using the multi-scale spatiotemporal feature fusion network MSFF-Net. Combined with a deep reinforcement learning agent and a multi-objective joint reward function, the training is optimized to generate a redirection walking control strategy, enabling users to explore safely and continuously in the virtual environment.
It enables precise perception and effective guidance for users in complex virtual-physical spaces, simultaneously optimizing exploration efficiency, obstacle avoidance effects, and user comfort, extending the continuous collision-free walking distance for users, and enhancing the immersion and continuity of VR exploration.
Smart Images
Figure CN122223281B_ABST