Deep Learning Positioning Using CSI-RS Magnitude Maps
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing positioning technologies, such as those based on GPS and cellular networks, face accuracy issues indoors and rely heavily on communication company data, limiting their effectiveness in various environments.
Innovation Solution
A deep learning-based method utilizing multi-channel or multi-antenna wireless signal data to generate a positioning database model by processing Channel State Information Reference Signals (CSI-RS) into magnitude map images, enabling precise location estimation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If GPS is used for positioning, then positioning availability is improved in wide open areas, but positioning accuracy deteriorates indoors and in downtown places
Solution Approach 1:
The patent introduces Wi-Fi signals as an intermediary positioning medium. Instead of relying solely on GPS satellites, the system uses Wi-Fi signal characteristics (signal strength, angle of arrival, time of arrival) as intermediaries to determine position indoors and in urban areas where GPS is unavailable or inaccurate.
Solution Approach 2:
The patent creates a universal positioning system that can operate in multiple environments (outdoor, indoor, downtown) by combining GPS, Wi-Fi, and cellular network positioning methods. The system selects the appropriate positioning method based on the current environment, making it universally applicable across different location scenarios.
2Adaptability or versatility
If cellular network-based positioning is used, then positioning can be performed without GPS, but positioning accuracy deteriorates because base station and repeater cover wide areas
Solution Approach 1:
The patent segments the positioning function by introducing multiple positioning methods (GPS, Wi-Fi, cellular) that can operate independently. Each method serves a specific purpose: GPS for outdoor positioning, Wi-Fi for indoor positioning, and cellular for general coverage. This segmentation allows the system to achieve both independence from GPS and maintained accuracy by selecting the appropriate method for each scenario.
3Measurement precision
If Wi-Fi-based indoor positioning is used, then positioning accuracy is improved, but time and expense increase due to signal collection requirements and AP deployment obstacles
Solution Approach 1:
The patent performs preliminary actions by pre-collecting and storing Wi-Fi signal data in a database before actual positioning is needed. The system pre-processes signal strength, angle of arrival, and time of arrival measurements and stores them along with corresponding location information. This preliminary data collection eliminates the need for real-time signal collection during positioning operations, significantly reducing time requirements.
4Measurement precision
If multi-channel and multi-antenna wireless signal data is used, then positioning accuracy is improved, but device complexity increases
Solution Approach 1:
The patent extracts only the necessary information from the complex multi-channel and multi-antenna wireless signal data. Instead of processing all raw signal data, the system extracts specific features such as signal strength, angle of arrival, and time of arrival from the CSI-RS measurements. This extraction approach maintains positioning accuracy while significantly reducing processing complexity by focusing only on the most relevant signal characteristics.
Data Source
AI summary
Disclosed herein are an apparatus and method for precise positioning based on deep learning. The method performed by the apparatus includes setting a collection location and a collection environment, collecting wireless signal data based on the collection location and the collection environment, generating a magnitude map image for training from the wireless signal data, and generating a positioning DB model by learning the image characteristics of the magnitude map image for training through deep-learning-based training.