Machine Learning Weighting for Angle-of-Arrival Location Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current location solutions in wireless communication networks, which use multiple access points and antennas to estimate the location of wireless devices, suffer from inaccuracies due to unequal contributions from all access points and antennas, with some contributing to degraded accuracy.
Innovation Solution
A machine learning model is developed to focus on particular location information from radio frequency elements by weighting or excluding certain access points and antennas based on their contribution to the location computation, using probability heatmaps and neural networks to determine the best combination of access points and assign weights to antennas for improved accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Area of stationary object
If multiple access points and antennas are used to estimate device location, then location coverage is improved, but location accuracy deteriorates due to unequal contributions from all elements
Solution Approach 1:
The patent applies local quality by assigning different weights to different antennas and access points based on their individual contribution to location accuracy. The machine learning model evaluates each RF element's quality metrics (signal strength, angle of arrival reliability, noise levels) and assigns localized weightings, allowing high-quality elements to contribute more while low-quality elements are downweighted or excluded. This resolves the contradiction by maintaining comprehensive coverage through multiple elements while ensuring only quality contributions affect the final location estimate.
Solution Approach 2:
The patent changes parameters by dynamically adjusting the weight coefficients of different access points and antennas based on real-time quality assessments. The machine learning model continuously evaluates signal characteristics and modifies the contribution parameters of each RF element. This allows the system to maintain broad coverage while adapting the effective weight of each element to optimize location accuracy under varying environmental conditions.
2Device complexity
If all access points and antennas contribute equally to location computation, then device complexity is reduced, but location accuracy deteriorates
Solution Approach 1:
The patent applies preliminary action by pre-evaluating and weighting the quality of each access point and antenna before performing location computation. The machine learning model performs offline training and quality assessment to establish weight coefficients that are then applied during runtime. This preliminary weighting reduces online computational complexity while maintaining high location accuracy, as the system only needs to apply pre-determined weights rather than performing complex equal-contribution calculations for all elements.
Solution Approach 2:
The patent extracts and removes low-quality access points and antennas from the location computation process through the machine learning model's quality assessment. By identifying and excluding elements that would degrade accuracy, the system reduces the number of elements requiring computation while maintaining or improving location precision. This extraction principle resolves the contradiction by eliminating unnecessary computational overhead from poor-quality elements.
3Measurement precision
If machine learning model weights or excludes certain access points and antennas, then location accuracy is improved, but device complexity increases
Solution Approach 1:
The patent applies self-service by implementing a machine learning model that autonomously evaluates the quality of each access point and antenna and automatically assigns weights or exclusions without requiring manual configuration. The system self-optimizes by continuously monitoring signal characteristics and adjusting element contributions based on observed performance. This automation improves location accuracy while managing complexity through self-adjustment rather than manual intervention.
Solution Approach 2:
The patent implements feedback mechanisms where the machine learning model continuously monitors the performance and quality of each RF element, using this feedback to dynamically adjust weighting coefficients. The system evaluates location estimation quality and signal characteristics, then feeds this information back to modify the contribution of individual elements. This feedback loop improves location accuracy by adapting to changing conditions while managing system complexity through automated closed-loop control.
Data Source
AI summary
In one embodiment, a device obtains a machine learning model indicative of how to focus on particular location information from a plurality of radio frequency (RF) elements to provide an accurate location estimate of a wireless client based at least in part on angle-of-arrival information of the wireless client. When the device then obtains location information regarding the wireless client from the plurality of RF elements, it may apply the machine learning model to the location information regarding the wireless client to focus on particular location information of the location information from the plurality of RF elements. The device may then estimate a physical location of the wireless client based on focusing on the particular location information during a locationing computation.