Indoor Skeletal Map Generation for Accurate Mobile Localization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for generating indoor space maps are time-consuming, labor-intensive, and prone to errors due to variability in floor plan formats, making satellite-based navigation unreliable and requiring manual conversion that lacks accuracy.
Innovation Solution
A hybrid approach combining mechanized generation with manual intervention using trained machine-learning models, such as CNN, GNN, or GAN, to create accurate indoor space maps by preprocessing raw digital representations, selecting regions of interest, and applying feature detection techniques.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If conventional manual conversion methods are used to generate indoor space maps, then accuracy can be maintained through human review, but the process becomes time-consuming and labor-intensive
Solution Approach 1:
The patent segments the map generation process into distinct modules: raw digital representation acquisition, preprocessing, machine learning model processing, and manual review. This segmentation allows automated processing of routine tasks while reserving manual intervention for quality assurance, thereby reducing overall time while maintaining accuracy.
Solution Approach 2:
The patent introduces machine learning models as an intermediary between automated map generation and manual review. These models process the raw digital representations and generate preliminary maps, acting as a bridge that handles the time-consuming processing while allowing human reviewers to focus on accuracy verification rather than complete map creation.
2Measurement precision
If satellite-based navigation systems are used for indoor positioning, then outdoor navigation accuracy is maintained, but the systems become unreliable indoors due to signal availability issues
Solution Approach 1:
The patent creates a digital copy of the indoor space environment through machine-generated maps that replicate the physical layout. This digital twin serves as a reference framework that enables positioning without satellite signals, effectively copying the spatial information needed for navigation in environments where satellite signals are unavailable.
3Productivity
If machine learning models are used to automatically generate indoor space maps, then generation speed and accuracy are improved, but the device complexity increases
Solution Approach 1:
The patent employs universal machine learning models that can process multiple types of raw digital representations (floor plans, CAD drawings, photographs) and generate standardized map outputs. This multi-functionality allows a single complex system to handle diverse input formats, reducing the need for multiple specialized systems and thereby managing complexity more effectively.
4Adaptability or versatility
If manual conversion of floor plans to machine-usable maps is performed, then format variability can be addressed through human judgment, but the ease of manufacture deteriorates due to labor requirements
Solution Approach 1:
The patent applies parameter changes by training machine learning models to recognize and adapt to various floor plan formats, styles, and conventions. The models learn to interpret different visual representations and convert them into standardized machine-usable maps, automatically adjusting to format variability without requiring manual conversion for each format type.
Data Source
AI summary
Examples of generating a map for an indoor space are described. In one example, a raw digital representation of the indoor space is obtained and preprocessed by at least one of marking a region-of-interest (ROI) in the raw digital representation and classifying the indoor space into a space-category to obtain a preprocessed representation. Using the preprocessed representation, a skeletal map is generated including indicators to identify elements of the indoor space. Revision inputs are received to modify an erroneous indicator from amongst the multiple indicators. The erroneous indicator is revised to a modified indicator and the modified indicator and unmodified indicators are used to create a modified skeletal map which is usable for localizing a mobile device in the indoor space.