Location-Aware Adaptation of Augmented Reality Narratives

Document Title

Location-Aware Adaptation of Augmented Reality Narratives

Document Information

• Subject Area: Scene adaptation techniques for augmented reality (AR) narratives
• Keywords: Interactive narrative, augmented reality, drama, path planning, scene adaptation, narrative generation, user interaction, human-computer interaction, location-aware technology, optimization algorithm

Research Background and Issues

• Identified Problems or Challenges:

  1. AR narratives require integrating virtual events and characters with physical locations in the real world, posing high demands on event-location matching.
  2. Current location-aware AR content creation often requires developers to manually assign narrative events to real-world maps, which is time-consuming and lacks scalability.
  3. The primary challenge lies in ensuring narrative events occur at real-world locations that match the surrounding environmental context to convey meaningful storylines.
  4. Reusing the same story content in different real-world environments necessitates multiple rounds of manual adjustments.

• Research Significance: This study aims to provide an automated, scalable solution for quickly adapting interactive narrative content to different physical spaces. Such technology simplifies the AR narrative design process while offering users a more immersive and coherent experience.

• Research Motivation: With the growing popularity of AR devices, efficiently and automatically adapting narrative content to real-world physical spaces has become a key technical requirement for AR storytelling.

• Related Work:

  1. Previous research primarily focused on virtual content generation, scene semantic matching, or narrative tool development in mixed augmented reality, with limited attention to overall narrative-real-world map matching.
  2. Macvean et al. proposed the WeQuest tool for manually assigning events, but it lacks automation.
  3. Existing studies like FLARE focus on micro-level layout generation, while this study targets high-level narrative adaptation issues.

Solution

• Proposed Solution: An optimization-based automated narrative adaptation method is proposed. This method designs a navigation graph generation mechanism for AR narrative content, assigning narrative events to compatible physical locations while considering factors such as travel distance and distribution uniformity.

• Innovative Features:

  1. Automated optimization of narrative event location assignments, integrating event compatibility with map region type matching.
  2. Generation of a navigation graph for players, guiding them to explore different branches of the narrative in the real world.
  3. Extended functionalities, such as event center design, landmark visibility constraints, and hard location constraints.

• Implementation Steps:

  1. Input: Includes a story tree and a real-world map.
  2. Representation:
     • The "story tree" represents the narrative structure, including all possible events and storyline branches.
     • The real-world map is annotated with region types (e.g., work areas, residential areas), and each event specifies its desired region type.
  3. Define optimization objectives:
     • Event location compatibility.
     • Walking distance.
     • Regularization cost for travel distribution uniformity.
  4. Optimization search:
     • Use the Markov Chain Monte Carlo (MCMC) method to search for the optimal event assignment scheme.
     • Generate new candidate schemes and evaluate the cost function until optimization converges.
  5. Output: A generated navigation graph that guides players through narrative events sequentially.

Research Outcomes

• Specific Outcomes:

  1. Successfully adapted and distributed interactive narrative content across various real-world map scenarios (e.g., university campuses, Tokyo International Forum, Hong Kong Convention Center).
  2. User studies indicated that narrative event-location matching based on optimization significantly improved the rationality of storylines and user experience.
  3. The method efficiently adapted narratives of varying scales (medium-scale with 20 events, large-scale with 38 events).
  4. Experiments demonstrated the method's extensibility to meet diverse design requirements (e.g., concentrated event distribution, safety constraints).

• Advantages:

  1. High degree of automation, saving significant manual effort.
  2. Achieves shorter total walking distances and more uniform event distribution, outperforming manual designs in efficiency.
  3. Easily extensible, allowing for the addition of design constraints based on practical needs.

• Experimental or Evaluation Results:

  1. User study results showed that considering "event-location compatibility" costs led to higher user experience scores (average 4.55), while disregarding it resulted in significantly lower scores (average 1.85).
  2. Compared to manual assignment methods, automated optimization generated superior event assignment schemes in a shorter time (approximately 5 minutes), whereas manual methods typically required around 37 minutes.

• Limitations and Future Directions:

  1. The current method relies on the completeness of input map region types, which may not cover all real-world scenarios.
  2. It does not yet perform real-time dynamic analysis of location compatibility, such as variations in active areas during different times of the day.
  3. Lacks support for low-level scene details (e.g., object interactions), limiting the adaptation of more complex virtual content.
  4. Potential risks related to safety and privacy, such as assigning events in traffic-heavy areas.

Future Work:

• Dynamically identify regional characteristics to enhance scene adaptation flexibility.
• Further explore the relationship between walking distance and user experience.
• Introduce real-time computation and wearable device sensor support to enhance narrative interactivity.
• Extend the algorithm to support multi-user, large-scale, and diverse narrative scenarios.