Probabilistic Signal Fusion for Multi-Modal Localization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing localization technologies face challenges in accurately determining the position of electronic devices in urban-canyon, semi-indoor, and deep indoor settings due to weak or unavailable GPS signals, and the limitations of combining multiple heterogeneous signals, such as WiFi, Bluetooth, and geomagnetism, which suffer from noise and varying sampling rates, leading to inaccuracies and user experience issues.
Innovation Solution
A probabilistic signal fusion system, SiFu, that dynamically combines an arbitrary combination of heterogeneous signals, including geolocation, WiFi, Bluetooth, 4G/5G communication, and inertial navigation signals, using a likelihood processor and particle filter to compute location likelihoods and adapt to changing signal conditions, allowing for signal addition or removal without retraining, and employing machine learning techniques like denoising autoencoders and dynamic time warping for robustness.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If multiple heterogeneous signals (WiFi, Bluetooth, geomagnetism) are combined for localization, then coverage and availability are improved, but noise and measurement errors increase
Solution Approach 1:
The patent applies parameter changes by dynamically adjusting the weighting parameters (α, β, γ) in the fused localization estimate equation based on signal availability and quality conditions. When GPS is available, it receives higher weight; when unavailable, WiFi and geomagnetism weights are adjusted accordingly. This allows the system to adapt to different environmental conditions while maintaining optimal localization accuracy despite combining multiple noisy signals
Solution Approach 2:
The patent creates a composite localization system that fuses multiple heterogeneous signal sources (GPS, WiFi, geomagnetism, inertial sensors) into a unified localization estimate. Each signal type contributes its strengths while the fusion algorithm mitigates individual weaknesses, similar to how composite materials combine different materials to achieve superior properties that individual materials cannot provide alone
2Productivity
If signal sampling rates are increased to improve localization frequency, then user experience is improved, but energy consumption increases
Solution Approach 1:
The patent applies dynamics by implementing adaptive sampling rates for different signal sources based on their characteristics and current availability. GPS is sampled at lower rates when available, WiFi sampling is adjusted based on access point detection, and geomagnetism sampling is tuned to provide sufficient updates without excessive frequency. This dynamic adjustment maintains adequate localization frequency while significantly reducing overall energy consumption compared to uniform high-rate sampling of all signals
3Measurement precision
If GPS is used for outdoor localization, then accuracy is improved, but availability deteriorates in urban-canyon and indoor settings
Solution Approach 1:
The patent applies segmentation by dividing the operational environment into distinct segments (outdoor open space, urban-canyon, indoor) and assigning different primary localization strategies to each segment. GPS is used as the primary source in outdoor segments where available, while WiFi and geomagnetism become primary in indoor/urban-canyon segments. This segmentation allows the system to maintain high accuracy in each specific environment while ensuring overall availability across all settings
4Adaptability or versatility
If RF signals (WiFi, Bluetooth) are used for indoor localization, then availability is improved, but noise and measurement errors increase
Solution Approach 1:
The patent uses geomagnetism as an intermediary signal that provides a stable reference framework for indoor localization. While WiFi and Bluetooth provide availability in indoor environments, they introduce noise and multipath errors. The geomagnetism signal, being relatively stable and omnidirectional, serves as an intermediary reference that helps filter and correct the noisier RF signals, thereby maintaining indoor availability while reducing the impact of RF signal noise on overall localization precision
Data Source
AI summary
The present invention provides a signal fusion system that is a probabilistic system fusing an arbitrary combination of heterogeneous signals for determining the location of electronic devices. The system includes a signal sampling device for detecting one or more signals emitted by an electronic device to be located, including geolocation signals, WiFi signals, Bluetooth signals, 4G communication signals, 5G communication signals, geomagnetism signals, or inertial navigation system signals (INS). A likelihood processor cooperates with the signal sampling device to receive information about selected sampled signals, and creates a grid of reference points for an interested area in which the electronic device may be located. The likelihood processor independently computes, for each selected sampled signal, a location likelihood that is a probability of observing the sampled signal given that the electronic device is located at different reference points in the grid.