Indoor Asset Location via ML Zone Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for locating assets in indoor environments, such as GPS and RF technologies, face challenges due to multipath interference from obstacles, resulting in low accuracy and high costs, especially in complex settings like warehouses.
Innovation Solution
A method for training a machine learning model using dynamic data collection, where training tags are moved within zones to generate training data, reducing the impact of multipath interference and simplifying data collection, allowing for precise asset location determination without the need for synchronized receivers or extensive tag deployment.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If GPS tracking is used to locate assets, then live tracking capability is provided, but accuracy deteriorates to >6m in indoor environments
Solution Approach 1:
The patent replaces GPS satellite-based positioning with an indoor RF positioning system using access points and mobile devices. The system substitutes the mechanical/satellite-based GPS approach with electromagnetic signal-based indoor positioning, achieving both live tracking and improved accuracy within indoor environments by using locally deployed infrastructure rather than satellite signals.
Solution Approach 2:
The patent changes the measurement parameters from GPS coordinates (accurate outdoors but poor indoors) to RF signal strength measurements (RSSI) combined with trilateration. By switching to different physical measurement principles suitable for indoor environments, the system achieves accurate location tracking where GPS fails.
2Ease of manufacture
If RF technology with trilateration is used to locate assets, then cost is reduced compared to GPS, but accuracy deteriorates due to multipath interference
Solution Approach 1:
The patent implements feedback mechanisms where mobile devices continuously measure RF signal strengths from multiple access points and update location estimates in real-time. The system uses feedback from signal strength variations to compensate for multipath effects and improve positioning accuracy, allowing cost-effective RF-based positioning to overcome its traditional accuracy limitations.
Solution Approach 2:
The patent transitions from static trilateration based on fixed access point locations to dynamic positioning that continuously adapts to changing RF conditions. By using real-time signal strength measurements and updating positions dynamically, the system compensates for multipath interference and achieves accurate tracking at lower cost.
3Measurement precision
If the number of receivers is increased to improve RF location accuracy, then measurement precision improves, but device complexity increases
Solution Approach 1:
The patent makes mobile devices serve multiple functions: they act as both the positioning target (tracking the asset) and the positioning instrument (measuring RF signals to determine location). This eliminates the need for separate fixed receivers throughout the environment, reducing system complexity while maintaining accuracy through the mobile device's own RF measurements.
Solution Approach 2:
The patent enables the mobile device to perform its own positioning by having it measure RF signal strengths from access points and compute its location independently. The mobile device serves itself as the positioning instrument, eliminating the need for extensive receiver infrastructure and reducing overall system complexity while maintaining measurement precision.
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 approach enables efficient and accurate location of assets in indoor environments by leveraging machine learning to determine the zone of an asset's location, reducing operational costs and complexity, and improving precision beyond traditional RF-based methods.
Implementation Method 1
Multipath interference occurs when an RF signal arrives at a receiver via two or more routes. This results in the total length of each signal path, and thus the time delay and phase of each received signal, to be different.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A method for training a machine learning model for locating an asset in an environment when the machine learning model is executed on a computer system, wherein the environment comprises a plurality of zones. The method comprises receiving data corresponding to a first set of training signals, wherein the first set of training signals are received at a plurality of receivers, within a first data collection period, from a plurality of training tags located in a first zone of the plurality of zones. The plurality of training tags are moved from a first position in the first zone to a second position in the first zone within the first data collection period. The method also comprises receiving data corresponding to a second set of training signals, wherein the second set of training signals are received at the plurality of receivers, within a second data collection period, from a plurality of training tags located in a second zone of the plurality of zones. The plurality of training tags are moved from a first position in the second zone to a second position in the second zone within the second data collection period. The method further comprises generating training data for the first zone, wherein the training data for the first zone comprises values of the training signals received from training tags located in the first zone within a first predefined deviation window, Wt, within the first data collection period, each value associated with the receiver at which a respective training signal was received. The method further comprises generating training data for the second zone, wherein the training data for the second zone comprises values of the training signals received from training tags located in the second zone within a second predefined deviation window, Wt', within the second data collection period, each value associated with the receiver at which a respective training signal was received. The method also comprises training, using the training data for the first zone and the training data for the second zone, a machine learning model to output a zone as a determined location of an asset in the environment based on an input including data corresponding to one or more signals received at one or more receivers from a tag associated with the asset.