Embodied Tentacle: Mapping Design to Control of Non-Analogous Body Parts with the Human Body

Document Title

Embodied Tentacle: Mapping Design to Control of Non-Analogous Body Parts with the Human Body

Document Information

• Subject Area: Human-Computer Interaction, Virtual Reality, Non-Humanoid Structure Mapping Design
• Keywords: Non-Humanoid Avatar, Virtual Reality, Gesture Interaction, Body Schema, Human Augmentation, Mapping Design, Multi-Joint Limb, User Experience, Virtual Arm Mapping, Embodiment

Research Background and Problem

• Research Background: Non-humanoid bodies (such as octopus tentacles or mechanical arms in virtual reality) hold significant potential in gaming, animation, and human augmentation. They not only expand physical capabilities but also create novel interactions and immersive experiences. However, due to the structural differences between the human body and these body parts, providing users with natural and comfortable control experiences poses a substantial challenge.

• Research Problem: How to design effective mapping strategies that enable humans to control non-humanoid bodies with completely different structures.

• Research Motivation: Although some studies have proposed mapping design methods (e.g., gesture-to-structure mapping or structural alignment), their applicability may be limited by human physiological constraints or specific usage purposes. Therefore, there is an urgent need to develop more universal and practical control strategies for non-humanoid bodies.

Solution

• Research Method:

  • Design a virtual octopus arm with 12 joints and conduct experiments using different mapping methods (e.g., sequential mapping or non-sequential mapping).
  • Investigate the impact of different mapping methods on user operation experience, including task performance, embodiment, and user preference.

• Innovations:

  • Identify three key factors influencing the mapping experience of non-humanoid bodies:
    1. Visual and Configurational Similarity: Maintaining visual and structural similarity between user finger movements and virtual limb joint movements.
    2. Kinematics Suitability for the User: Considering the characteristics of human joint movements, as not all fingers can move with high independence and precision.
    3. Correspondence with Everyday Actions: Whether users can transfer their habitual movements to control the virtual limb.

• Implementation Steps:

  1. Provide four mapping conditions: Non-Sequential, Sequential-Index-Based, Sequential-Little-Based, and Sequential-Across.
  2. Design two tasks: Reaching Task (dynamic smoothness) and Pose Imitation Task (accuracy).
  3. Collect subjective and objective data through experiments, including success rate, completion time, posture error, and user preference.
  4. Use questionnaires and interviews to understand users' embodiment and overall experience.

Research Results

• Specific Findings:

  1. In terms of task performance, Sequential-Index-Based mapping showed significant operational advantages, while Sequential-Across mapping enhanced users' sense of agency.
  2. Comparing performance across the four mapping conditions revealed that sequential mapping significantly improved user control comfort and operational accuracy.

• Advantages Compared to Existing Solutions: This study systematically defines various key factors affecting control experiences, surpassing previous mapping designs that were simply based on gesture pairing or structural replication.

• Experimental Results:

  • In the Reaching Task, Sequential-Index-Based mapping had significantly shorter completion times compared to Sequential-Across.
  • In the Pose Imitation Task, Non-Sequential mapping performed the worst, while Sequential-Index-Based mapping received the highest scores in user preference and usage experience.
  • Users showed a preference for sequential mapping, especially Sequential-Index-Based, which aligns with everyday habits.

• Limitations and Future Directions:

  1. Limitations:

     • The experiment duration was relatively short, preventing exploration of long-term learning effects.
     • The study was limited to one-to-one joint mapping and did not explore more complex high-degree-of-freedom control.

  2. Future Directions:

     • Conduct long-term user studies to evaluate learning curves and performance differences after task saturation.
     • Design more complex mapping strategies (e.g., nonlinear mappings based on machine learning).
     • Extend research to a wider variety of non-humanoid structures (e.g., full-body multi-limb control).
     • Incorporate physiological measurement methods (e.g., skin conductance) to further investigate the mechanisms of embodiment.