Data Association Verification via Linear Inequality Reframing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current navigation and detect and avoid systems rely on data association algorithms that are challenging to verify correctly due to their non-linear nature, making it difficult to ensure the accuracy of feature and target identification using formal methods-based verification.
Innovation Solution
A system that includes a processor and non-cooperative sensors, featuring a correlator module to output correlated tracks and a verification module that reframes data association into linear inequalities, allowing for formal methods-based verification to ensure the correctness of the tracks by comparing properties within sets a and b using branching and bounding techniques.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If formal methods-based verification is applied to data association algorithms, then verification rigor is improved, but the non-linear nature of the algorithms makes them unsuitable for formal verification without substantial abstraction
Solution Approach 1:
The patent transforms the verification problem by changing the parameter representation from non-linear statistical properties to linear inequality constraints. By formulating data association verification as a system of linear inequalities rather than attempting direct verification of non-linear algorithms, the patent enables formal methods to be applied without requiring substantial abstraction of the original algorithms.
Solution Approach 2:
The patent introduces an intermediary formulation layer that translates the non-linear data association problem into a linear inequality framework. This intermediary representation serves as a bridge between the complex non-linear algorithms and the formal verification methods, allowing verification to proceed on the linearized representation while maintaining fidelity to the original problem.
2Ease of operation
If testing and simulation methods are used to gain confidence in data association implementations, then ease of operation is improved, but these methods are insufficient to guarantee the correctness of the association
Solution Approach 1:
The patent replaces the mechanical testing and simulation approach with a formal mathematical verification system. Instead of relying on empirical testing methods, the patent substitutes a rigorous formal methods-based verification framework that provides mathematical guarantees of correctness, thereby eliminating the insufficiency of testing while maintaining ease of operation through systematic verification procedures.
3Reliability
If prior research efforts define abstractions of data association algorithms in formal notations and prove high-level properties using theorem provers, then verification rigor is improved, but these approaches are not scalable to analyze full-scale models without abstractions
Solution Approach 1:
The patent applies partial action by focusing verification efforts on the critical linear inequality constraints rather than attempting to verify every aspect of the full-scale data association model. By verifying the essential linear constraints that govern data association correctness, the patent achieves scalable verification without requiring complete abstraction of the entire system.
Solution Approach 2:
The patent segments the verification problem into distinct linear inequality components that can be analyzed independently. By dividing the complex data association verification into manageable linear constraint segments, the patent enables scalable analysis of full-scale models while maintaining verification rigor through systematic examination of each segment.
Data Source
AI summary
A method for providing assurance of data association comprises formulating a track permutation matrix of measurement statistics based on sensor measurement returns; receiving a first set of correlated tracks from a correlator module (set a), identified as originating from a target; selecting a second set of correlated tracks from the track permutation matrix (set b), based on branching and bounding techniques; reframing verification of a data association framework into a linear inequality, with a first portion of the linear inequality based on set a, and a second portion of the linear inequality based on set b; and performing a formal methods-based verification procedure to determine whether properties of set a are bounded by properties of set b. If properties of the correlated tracks within set a are bounded by, or are less than or equal to, properties of the correlated tracks within set b, then the correlated tracks are verified.