Scene Model Generation via Ray Casting and Caching

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing artificial reality systems face challenges in efficiently generating, querying, and managing scene models, particularly in handling localization failures that lead to relocalization issues and frustration for users.

Innovation Solution

The system introduces a scene model that is an objective, system-managed representation of a user's physical environment, composed of anchors that describe static geometry and semantics. This scene model is generated using a scene capture workflow, and a scene realignment solution is provided to mitigate localization failures by relying on user input to align cached room data with the current map.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If the system uses automated scene capture workflows to generate scene models, then productivity is improved, but reliability deteriorates due to localization failures and relocalization issues

Engineering Contradiction:
Improvescene model generation efficiencyVSAvoidlocalization accuracy
Core Design Contradiction:
ProductivityVSReliability

Solution Approach 1:

The system performs preliminary scene capture and caching during initial setup or when the environment is stable. By capturing and storing scene data in advance (creating a cached scene model), the system prepares reliable reference data before potential localization failures occur, enabling faster recovery without re-scanning the entire environment.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The system creates and maintains cached copies of scene models that can be reused when the current localization fails. Instead of relying solely on real-time sensor data, the system references previously captured and stored scene representations, allowing it to recover from localization errors without requiring complete re-capture of the environment.

Inventive Principle:
Principle #26Copying

2Measurement precision

If the system implements comprehensive scene capture workflows to capture all environment details, then measurement precision is improved, but loss of time increases due to extensive re-scanning when relocalization is needed

Engineering Contradiction:
Improveenvironment representation accuracyVSAvoidre-scanning duration
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The system performs comprehensive environment capture and stores detailed scene models in cache during initial setup. This preliminary action ensures high measurement precision is achieved once, and the cached data can be reused multiple times without requiring re-scanning, thus eliminating the time loss that would otherwise occur during relocalization events.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

When localization failures occur, the system recovers by referencing previously captured and cached scene data instead of discarding the original comprehensive capture effort. The cached representations serve as recoverable copies that maintain high measurement precision without requiring time-consuming re-scanning of the entire environment.

Inventive Principle:
Principle #34Discarding and recovering

3Object-affected harmful factors

If the system manages detailed scene models with multiple anchors and components, then object-affected harmful factors are reduced through better environment understanding, but device complexity increases

Engineering Contradiction:
Improvelocalization failure impactVSAvoidscene model structure
Core Design Contradiction:
Object-affected harmful factorsVSDevice complexity

Solution Approach 1:

The scene model is segmented into discrete anchors, each representing specific environmental features or objects. This segmentation allows the system to manage complexity by breaking down the overall scene into manageable components, where each anchor can be independently captured, stored, and referenced, reducing the impact of localization failures on the entire scene.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The system maintains cached copies of the segmented scene model structure, allowing it to recover from localization failures by referencing previously stored anchor data. This approach manages the complexity of detailed scene models by providing a recoverable snapshot that can be restored without re-capturing the entire complex structure.

Inventive Principle:
Principle #34Discarding and recovering

Data Source

PatentUS12307575B2Scene capture via artificial reality systems
Publication Date: 2025.05.20 META PLATFORMS TECHNOLOGIES LLC
  • US12307575B2 patent drawing
  • US12307575B2 patent drawing
  • US12307575B2 patent drawing

AI summary

In particular embodiments, a computing system may initiate a scene capture process to capture a scene. The scene may include one or more of planes or objects. The system may send a first set of instructions to outline one or more planes of the scene. The system may cast a first set of rays to outline the one or more planes. The system may create the one or more planes based on the first set of rays. The system may send a second set of instructions to outline one or more objects of the scene. The system may cast a second set of rays to outline the one or more objects. The system may create the one or more objects based on the second set of rays. The system may generate a scene model of the scene based on the one or more planes and the one or more objects.