Indoor Wireless Positioning Using XAI-Selected CIR Features
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional wireless positioning technologies face challenges in achieving accurate indoor positioning due to multipath signal propagation, requiring significant computational resources and being unsuitable for low-power mobile devices, and lack clarity in the relationship between learning models and accuracy results.
Innovation Solution
A system and method utilizing a lightweight neural network architecture that includes pre-processing of CIR data into a two-dimensional image, applying explainable artificial intelligence (XAI) to select critical positioning data, and training auto-encoders and second neural networks to minimize computational requirements while maintaining high accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If deep learning technology is used to select peaks from CIR data, then measurement precision is improved, but device complexity and computational requirements increase significantly
Solution Approach 1:
The patent segments the deep learning model into two distinct networks: a first deep learning network that processes all CIR data to identify potential peaks, and a second deep learning network that selectively processes only the identified peaks to determine line-of-sight status. This segmentation reduces the computational burden on the second network while maintaining high measurement precision through selective detailed analysis.
Solution Approach 2:
The patent applies partial action by using the first deep learning network to perform preliminary screening of all CIR data to identify candidate peaks, and then applying the more computationally intensive second deep learning network only to these selected candidates rather than the entire dataset. This approach achieves high precision where needed while reducing overall computational complexity.
2Measurement precision
If a comprehensive neural network is used to process all CIR data, then measurement precision is improved, but use of energy by moving object increases
Solution Approach 1:
The patent divides the neural network processing into two stages with different energy requirements: the first network performs low-energy preliminary peak identification on all CIR data, and the second network performs high-energy detailed analysis only on selected peaks. This segmentation reduces overall energy consumption while maintaining positioning accuracy.
Solution Approach 2:
The system applies full computational power selectively only when needed - the second energy-intensive neural network is activated only for peaks identified by the first network, rather than processing all CIR data with high computational resources. This partial application of intensive processing reduces energy consumption while preserving measurement precision.
3Reliability
If explainable AI is applied to analyze the neural network, then reliability is improved through clarity of causal relationship, but device complexity increases
Solution Approach 1:
The patent introduces an explanation generation unit that acts as an intermediary between the dual neural networks and the final positioning output. This unit generates human-interpretable explanations for the network decisions, improving reliability and model interpretability while keeping the core neural network structures relatively simple and modular.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
The solution achieves high-accuracy wireless positioning with reduced computational demands, enabling efficient operation on low-power mobile devices and cost savings by simplifying the neural network structure using explainable AI techniques.
Implementation Method 1
This wireless positioning technology performs ranging by measuring Time of Flight (ToF). In more detail, messages are exchanged between anchors and tags through wireless communication, time is measured, and distances are estimated using a speed of light.
Implementation Method 2
The conventional wireless positioning technology measures time using channel impulse response (CIR) data collected during wireless communication.
Data Source
AI summary
The present disclosure provides a system for wireless positioning system including a collection unit for collecting a plurality of CIR data and distance data between a plurality of anchor terminals and the mobile terminal, when the mobile terminal communicates with the plurality of anchor terminals installed in a space for performing positioning of the mobile terminal; a pre-processing unit for generating a plurality of pre-proceed data by pre-processing the plurality of CIR data; a first learning unit for learning a first artificial neural network based on the plurality of pre-processed data and the distance data; an analysis unit for analyzing the first artificial neural network to select a plurality of positioning critical data used for positioning of the mobile terminal among the plurality of pre-processed data; and a second learning unit for learning a second artificial neural network based on the plurality of positioning critical data and the distance data.