Unsupervised Neural Network Localization Using Channel State Information
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Next-generation wireless networks face challenges in localization and environment mapping due to the data-driven nature of neural networks, which are computationally expensive and difficult to train, especially with unlabeled CSI data from base stations and user equipment.
Innovation Solution
An unsupervised learning method using artificial neural networks that processes channel state information (CSI) to determine user equipment locations and model propagation environments, employing auto-encoders and differentiable channel models to infer locations and reconstruct CSI values without labeled data, leveraging multi-path techniques for improved localization and mapping.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If neural networks are used for localization and mapping, then localization accuracy and environment mapping capability are improved, but computational cost and training difficulty increase
Solution Approach 1:
The system performs self-training by using the neural network to generate its own training data through unsupervised learning on CSI measurements. The network learns to map CSI to locations and environments without external labeled data, automatically improving its own performance through iterative training on the wireless channel characteristics.
Solution Approach 2:
The patent transforms the training approach by changing from supervised learning with labeled data to unsupervised learning using raw CSI parameters. The neural network learns to extract location and environment information directly from CSI measurements, changing the parameter representation from labeled locations to raw signal characteristics.
2Measurement precision
If labeled data is used for training neural networks, then training accuracy is improved, but data availability and labeling cost deteriorate
Solution Approach 1:
The system generates its own training data by processing unlabeled CSI measurements through the neural network. The network learns to create meaningful training examples from raw wireless channel data, eliminating the need for external labeling resources and enabling self-sufficient training.
Solution Approach 2:
The patent creates synthetic training data by copying and transforming raw CSI measurements into labeled equivalents through the neural network's learned representations. This allows the system to generate abundant training data from limited unlabeled measurements, effectively copying the structure of labeled data without requiring actual labeled examples.
3Area of stationary object
If more base stations and sensors are deployed, then localization coverage and mapping detail are improved, but system complexity and cost increase
Solution Approach 1:
The neural network performs multiple functions simultaneously: it processes CSI measurements for localization, extracts environment information for mapping, and generates training data for continuous improvement. This multi-functionality reduces the need for separate specialized components and simplifies the overall system architecture.
Solution Approach 2:
The patent combines localization and mapping functions into a single neural network model that processes CSI data to simultaneously determine user location and environmental characteristics. This merging of functions reduces system complexity compared to separate localization and mapping systems.
Data Source
AI summary
A method of training an artificial neural network (ANN), receives, from a base station, signal information for a radio frequency signal between the base station and a user equipment (UE). The artificial neural network is trained to determine a location of the UE and to map the environment based on the received signal information and in the absence of labeled data.