Machine Learning Path Generation for Virtual Space Narratives
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In virtual spaces, automated path generation for AI characters often leads to uninteresting locations, lacking narrative appeal, and fails to incorporate visually arresting scenes, making it challenging to create engaging cinematic experiences without pre-scripting.
Innovation Solution
A machine learning-based system that generates paths by considering the appeal of environments within the virtual space, using tagged points of interest and narrative attributes, allowing for the creation of unique and interesting paths that respond to user choices and interactions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If automated path generation is used for AI characters, then rendering efficiency is improved, but the paths lead to uninteresting locations from a story perspective
Solution Approach 1:
The system pre-computes and stores narrative attributes, points of interest, and path quality scores for all possible paths before actual rendering. This preliminary action allows the system to quickly retrieve and evaluate pre-analyzed path options during runtime, maintaining both efficiency and narrative quality without re-computing narrative attributes on the fly.
Solution Approach 2:
The patent replaces traditional manual path scripting with an automated machine learning-based system that uses narrative attributes and points of interest to generate paths. This substitution maintains narrative quality by using learned patterns of interesting locations while achieving automation and efficiency through algorithmic path generation rather than hand-crafted scripts.
2Adaptability or versatility
If procedural content generation is used, then adaptability to unfamiliar virtual spaces is improved, but narrative quality and visual appeal deteriorate
Solution Approach 1:
The system uses a machine learning training loop where path quality is evaluated based on narrative attributes and points of interest. The feedback from this evaluation is used to train and improve the path generation model, enabling it to learn what constitutes narratively interesting paths in diverse virtual spaces while maintaining adaptability to unfamiliar environments through continuous learning.
Solution Approach 2:
The patent changes the parameters used for path generation from simple geometric or navigational criteria to include narrative attributes such as points of interest, story relevance, and visual appeal. By incorporating these additional parameters into the path generation process, the system maintains narrative quality while adapting to procedural and unfamiliar virtual spaces.
3Extent of automation
If paths are generated without pre-scripting, then automation is improved, but paths fail to incorporate visually arresting scenes
Solution Approach 1:
The system pre-identifies and tags points of interest with visual appeal attributes before path generation. This preliminary tagging allows the automated path generation algorithm to automatically incorporate visually arresting scenes by selecting paths that pass through pre-identified interesting locations, maintaining both automation and visual appeal without manual scripting.
Solution Approach 2:
The patent replaces manual path scripting with an automated machine learning system that uses narrative attributes and points of interest to guide path generation. This substitution maintains visual appeal by using learned patterns of interesting locations while achieving full automation through algorithmic decision-making based on pre-computed narrative data.
Data Source
AI summary
The systems and methods presented herein are related to providing narrative experiences for users of a virtual space. A machine learning based technique may be implemented for generating paths through the virtual space that present the narrative experiences. A path may be based on tagged points of interest, as illustrated in stored narrative information associated with the virtual space. A machine-learning training loop may be applied wherein the system may be trained to recognize and/or understand what sorts of virtual content within the virtual space may be considered to be of narrative interest to audience members.