Wi-Fi Localization via Path Loss Estimation and EM Algorithm
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing indoor localization methods using received signal strength (RSS) measurements require training and specialized hardware, making them costly and complex, and are unreliable due to dynamic changes in indoor environments.
Innovation Solution
A method for unsupervised localization using a conventional Wi-Fi chipset that estimates path loss coefficients and device location through a log-distance path loss model, employing expectation maximization (EM) and non-linear Kalman filtering, without the need for initial training or hardware modifications, leveraging inertial measurement units (IMUs) for location change estimation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If specialized hardware is installed in the environment for indoor localization, then localization accuracy is improved, but device complexity and cost increase
Solution Approach 1:
The system uses existing Wi-Fi access points and conventional Wi-Fi chipsets to perform localization without requiring specialized hardware installation. The environment's existing infrastructure serves the localization function, eliminating the need for additional specialized devices while maintaining acceptable localization accuracy
Solution Approach 2:
Conventional Wi-Fi chipsets, originally designed for communication purposes, are made to serve dual functions: wireless communication and localization. This multi-functionality eliminates the need for specialized localization hardware while leveraging existing ubiquitous Wi-Fi infrastructure
2Measurement precision
If training is performed by measuring RSS levels offline in the indoor environment, then localization accuracy is improved, but reliability deteriorates due to dynamic environmental changes
Solution Approach 1:
The system performs online localization calculations using real-time RSS measurements and the log-distance path loss model, allowing the localization estimates to adapt dynamically to current environmental conditions rather than relying on static offline training data that becomes outdated when the environment changes
Solution Approach 2:
The system uses the log-distance path loss model with path loss coefficients that can be estimated online, allowing the localization parameters to adjust to changing environmental conditions such as occupancy changes, furniture movements, and AP location changes, thereby maintaining reliability in dynamic environments
3Reliability
If training is repeated whenever the environment changes, then localization reliability is improved, but loss of time increases
Solution Approach 1:
The system performs continuous online localization calculations using real-time RSS measurements and the log-distance path loss model, eliminating the need to stop and repeat training whenever environmental conditions change. The localization process continuously adapts to current conditions without interruption or retraining
4Measurement precision
If a large number of access points are deployed to support estimation of large number of parameters, then localization accuracy is improved, but device complexity and cost increase
Solution Approach 1:
The system uses the log-distance path loss model which reduces the number of parameters that need to be estimated compared to other models. By modeling path loss as a function of distance with a path loss coefficient, the system can achieve accurate localization with fewer access points, as each AP contributes efficiently to the parameter estimation without requiring a large deployed infrastructure
Data Source
AI summary
A method and system determines a location of a device by first measuring, in a receiver of the device, received signal strength (RSS) levels of reference signals transmitted by a set of access points (APs) arranged in an enclosed environment. An expectation maximization procedure is initialized with a uniform distribution of path loss coefficients of the set of APs. An expectation of a joint log-likelihood of the RSS levels is evaluated with respect to the uniform probability distribution of the path loss coefficients, and then updated iteratively until a termination condition is satisfied to produce final path loss coefficients. The location of the device is then based on the RSS levels and the final path loss coefficients.