Transferable Virtual-Physical Environmental Alignment With Redirected Walking.

Area of Science:

• Virtual Reality
• Human-Computer Interaction
• Robotics

Background:

• Redirected walking (RW) aims to create natural locomotion in virtual environments (VEs).
• Aligning virtual and physical environments is crucial for RW, enabling interactions like touching virtual objects that correspond to real-world counterparts.
• Existing methods face challenges when multiple virtual and physical target positions are transferable, complicating alignment.

Purpose of the Study:

• To address the challenge of virtual-physical environmental alignment at multiple transferable target positions in RW.
• To introduce a novel reinforcement learning (RL)-based method for improved alignment.
• To enhance the naturalness and immersion of walking in VEs.

Main Methods:

• Development of a novel RL-based redirected walking method.
• Design of a comprehensive reward function for dynamic virtual-physical target matching.
• Updating virtual target weights for reward computation based on alignment success.

Main Results:

• The proposed RL method achieved reduced physical distance error in environmental alignment compared to state-of-the-art techniques.
• The method required fewer resets, indicating more stable and continuous virtual locomotion.
• Simulated experiments and real user tests validated the effectiveness of the approach.

Conclusions:

• The novel RL-based RW method effectively solves the virtual-physical environmental alignment problem at multiple transferable target positions.
• This approach significantly improves the accuracy and efficiency of environmental alignment in RW.
• The findings contribute to more immersive and natural user experiences in virtual environments.