Learning-Based Safety Relevance Determination for Autonomous Software

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Manual change impact analysis for safety-critical software in autonomous systems is time-consuming, error-prone, and costly, leading to potential safety risks and system downtime due to the need for extensive personnel involvement and delayed software updates.

Innovation Solution

A learning-based method using a neural network to determine the safety relevance of software changes autonomously, reducing personnel requirements and errors by applying machine learning to input data records, including source code changes, and providing a probability-based indication of safety impact.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If manual change impact analysis is performed according to safety standards, then reliability of safety determination is improved, but productivity and time efficiency deteriorate

Engineering Contradiction:

Improvesafety determination reliabilityVSAvoidchange analysis productivity

Core Design Contradiction:

ReliabilityVSProductivity

Solution Approach 1:

The patent replaces the manual mechanical analysis process with an automated computer-based system that uses machine learning models to perform change impact analysis, thereby maintaining reliability while dramatically improving productivity

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The system enables automated self-analysis of software changes by using trained machine learning models that automatically determine safety relevance without requiring manual human intervention for each change assessment

Inventive Principle:

Principle #25Self-service

2Measurement precision

If extensive manual analysis is performed on hundreds of changes, then measurement precision of safety impact is improved, but loss of time increases

Engineering Contradiction:

Improvesafety impact precisionVSAvoidanalysis time

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The system performs preliminary training of machine learning models on historical change data with known safety outcomes, enabling rapid automated assessment of new changes without time-consuming manual analysis while maintaining precision through pre-learned patterns

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The system transforms the analysis from a time-intensive manual process into an automated process by changing the parameters of analysis through machine learning models that can rapidly evaluate multiple changes simultaneously with consistent precision

Inventive Principle:

Principle #35Parameter changes

3Reliability

If manual change impact analysis is performed, then reliability of safety determination is improved, but loss of substance in terms of personnel resources increases

Engineering Contradiction:

Improvesafety determination reliabilityVSAvoidpersonnel resources

Core Design Contradiction:

ReliabilityVSLoss of substance

Solution Approach 1:

The patent substitutes human personnel resources with an automated computer-based machine learning system that performs change impact analysis, thereby maintaining reliable safety determination while eliminating the need for extensive manual personnel involvement

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The system provides automated self-service capability for safety assessment, eliminating dependency on human analysts and reducing personnel resource consumption while maintaining determination reliability through consistent automated evaluation

Inventive Principle:

Principle #25Self-service

4Reliability

If software changes are delayed for manual analysis, then reliability of safety check is improved, but duration of system downtime increases

Engineering Contradiction:

Improvesafety check reliabilityVSAvoidsystem downtime

Core Design Contradiction:

ReliabilityVSDuration of action of moving object

Solution Approach 1:

The patent replaces slow manual safety analysis with automated machine learning-based analysis, enabling rapid safety verification that reduces system downtime while maintaining reliable safety checks through consistent automated evaluation

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The system enables continuous operation by performing rapid automated safety analysis that allows software changes to be implemented without prolonged system downtime, maintaining the continuity of useful action through efficient automated processing

Inventive Principle:

Principle #20Continuity of useful action

Data Source

PatentUS12073316B2Method for determining at least one indication of at least one change

Publication Date: 2024.08.27 SIEMENS AG

AI summary

Provided is a method for determining at least one indication of at least one change, having the steps of receiving at least one input data record having the at least one change and associated data, and determining the at least one indication of the at least one change by applying a learning-based approach to the at least one received input data record. The invention is also directed to a determination unit and a computer program product.