Multi-Modal Localization for GPS-Denied Environments
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing localization technologies face challenges in GPS-denied or degraded environments, such as urban canyons, mountainous regions, and celestial bodies, due to poor signal availability, multipath issues, excessive computation, and power consumption, and are unreliable in areas without satellite signal availability.
Innovation Solution
A Visual Positioning System (VPS) leveraging computer vision algorithms, 2D and 3D geospatial datasets, and image-based matching techniques, integrating multi-modal fusion with inertial measurement units (IMU), altimeters, and semantic scene understanding for precise geolocation, using skyline, landmark, and mesh VPS methods.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If GPS and satellite-based localization are used, then localization accuracy is improved, but reliability deteriorates in GPS-denied or degraded environments
Solution Approach 1:
The patent introduces an intermediary system that uses visual landmarks, sky features, and inertial sensors as mediators between the device and the environment for localization when GPS is unavailable. The system captures images of landmarks and sky features, extracts features and descriptors, matches them against pre-stored data, and computes position without requiring satellite signals.
Solution Approach 2:
The patent replaces the electromagnetic signal-based GPS system with an optical-mechanical vision-based localization system. Instead of using radio waves from satellites, the system uses camera-based image capture, feature extraction, and visual matching to determine position, substituting electromagnetic navigation with optical sensing and computational geometry.
2Reliability
If radio-frequency terrestrial beacons are used, then localization is possible in urban areas, but signal availability deteriorates due to multipath and jamming
Solution Approach 1:
The patent substitutes radio-frequency signal-based localization with optical vision-based localization. Instead of relying on RF beacons that suffer from multipath and jamming, the system uses camera images of landmarks and sky features, extracting visual descriptors and matching them against pre-stored data to determine position, thereby eliminating dependence on vulnerable RF signals.
Solution Approach 2:
The patent introduces visual features from images as intermediaries for localization. By capturing and analyzing images of landmarks, buildings, and sky features, the system creates a visual intermediary representation that can be matched against pre-stored environmental data to determine position without relying on direct RF signal transmission.
3Measurement precision
If image generation and comparison are performed for localization, then positioning precision is improved, but computation requirements increase excessively
Solution Approach 1:
The patent performs preliminary action by pre-generating and storing images or feature descriptors of landmarks and sky features at multiple possible positions before actual localization is needed. During localization, the system only needs to extract features from current images and match them against the pre-stored data, significantly reducing real-time computational requirements compared to generating and comparing images for every possible position.
Solution Approach 2:
The patent uses copying by creating a database of pre-computed images or feature descriptors representing environmental features at known positions. Instead of performing complex image generation and comparison for all possible positions during localization, the system copies pre-computed reference data and performs simpler feature matching against this stored information, reducing computational burden while maintaining precision.
Data Source
AI summary
A method for localizing a device includes determining a first position estimate for the device using acquired skyline data and determining second position estimate using an object model representing expected representing locations of instances of objects from multiple object classes. A combined position estimate is then determined based at least in part on the first and second position estimates.