Machine-Learning Validation of Satellite Integer Ambiguities

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing integer ambiguity validation methods in wireless communication systems, particularly in 5G networks, face increasing complexity and challenges in maintaining thresholds for logical tests, leading to inefficiencies in determining the correctness of integer ambiguity vectors for precise positioning.

Innovation Solution

Implementing a machine learning-based approach to validate integer ambiguity vectors by training a model to predict the probability of correctness, comparing it against a threshold to determine if the vector is correct, thereby reducing complexity and improving accuracy.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If traditional logical test methods are used for integer ambiguity validation, then the validation process can be implemented, but the system complexity increases and threshold maintenance becomes challenging

Engineering Contradiction:

Improveinteger ambiguity validation accuracyVSAvoidvalidation method complexity

Core Design Contradiction:

ReliabilityVSDevice complexity

Solution Approach 1:

The patent replaces traditional mechanical/logical test-based validation methods with a machine learning-based probabilistic model. The ML model processes satellite signal features and directly outputs a probability score indicating the likelihood of integer ambiguity vector correctness, eliminating the need for complex logical tests and threshold maintenance while improving validation reliability

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The patent transforms the validation approach by changing from deterministic logical tests to probabilistic scoring. The system outputs a probability value (0-1) representing the confidence level of the integer ambiguity vector, allowing for more flexible and accurate validation decisions based on the specific characteristics of each measurement scenario

Inventive Principle:

Principle #35Parameter changes

2Reliability

If more logical tests are added to improve validation accuracy, then reliability increases, but the difficulty of maintaining thresholds increases

Engineering Contradiction:

Improvevalidation accuracyVSAvoidthreshold maintenance ease

Core Design Contradiction:

ReliabilityVSEase of operation

Solution Approach 1:

The patent substitutes multiple logical tests with a single unified machine learning model that inherently learns the optimal decision boundaries during training. The model processes input features and directly outputs a probability score, eliminating the need for operators to maintain multiple logical test thresholds while preserving high validation accuracy

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

3Productivity

If traditional validation methods are used, then the process can be implemented, but the balance between fixing rate and incorrect fix rate is suboptimal

Engineering Contradiction:

Improvefixing rateVSAvoidincorrect fix rate

Core Design Contradiction:

ProductivityVSReliability

Solution Approach 1:

The patent changes the validation output from a binary correct/incorrect determination to a continuous probability score. This allows the system to dynamically adjust the decision threshold based on the desired balance between fixing rate and incorrect fix rate, optimizing both productivity and reliability for different operational requirements

Inventive Principle:

Principle #35Parameter changes

Solution Approach 2:

The machine learning model is trained using feedback from labeled training data, learning to distinguish between correct and incorrect integer ambiguity vectors. During operation, the model provides probabilistic feedback that can be used to adjust validation thresholds and optimize the balance between fixing rate and incorrect fix rate based on performance monitoring

Inventive Principle:

Principle #23Feedback

Data Source

PatentUS12422566B2Integer ambiguity validation with machine learning

Publication Date: 2025.09.23 QUALCOMM INC

AI summary

An integer ambiguity validation method includes: obtaining, at an apparatus in conjunction with the one or more receivers, a plurality of feature values that are based on satellite signals received by a mobile device; determining, at the apparatus, an integer ambiguity vector indicative of integer numbers of carrier phase cycles of the satellite signals between the apparatus and respective satellites; determining, at the apparatus, a probability of the integer ambiguity vector being correct by using the integer ambiguity vector in a machine learning algorithm; and determining, at the apparatus, whether the integer ambiguity vector is correct based on the probability of the integer ambiguity vector being correct and an integer ambiguity vector probability threshold.