Virtual Test Result Approximation for Autonomous Driving Validation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for verifying and validating autonomous vehicle functions require extensive time and resources, as they necessitate actual vehicle operation to determine critical test cases, which is inefficient for scenario-based testing in highly automated driving.
Innovation Solution
A computer-implemented method using an artificial neural network to approximate critical test results by defining a state space with driving situation parameters, employing reinforcement learning and Q-learning principles to identify subsets of test results, thereby reducing the need for extensive testing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If actual vehicle operation is used to determine critical test cases, then the accuracy and reliability of test results is improved, but the time and resource expenditure increases significantly
Solution Approach 1:
The patent creates a virtual copy of the vehicle system through artificial neural networks that replicate the behavior and decision-making processes of the actual autonomous driving system. This virtual model allows testing without physical vehicle operation, maintaining reliability while reducing time and resource costs.
Solution Approach 2:
The patent replaces the mechanical system of actual vehicle operation with a computational system using artificial neural networks and virtual simulations. This substitution maintains the ability to obtain reliable test results while eliminating the time-consuming and resource-intensive nature of physical testing.
2Reliability
If a large number of parameter combinations are tested, then the completeness of validation is improved, but the complexity and cost of testing increases
Solution Approach 1:
The patent applies partial action by using artificial neural networks to identify and prioritize only the most critical parameter combinations that need testing. Instead of exhaustively testing all possible combinations, the system focuses on the subset of test cases that are most likely to reveal safety issues, reducing complexity while maintaining validation completeness.
Solution Approach 2:
The patent transforms the testing approach by changing parameters from exhaustive enumeration to intelligent selection based on neural network analysis. The system dynamically adjusts which parameter combinations are tested based on learned patterns from training data, reducing the number of test cases needed while maintaining comprehensive validation.
3Reliability
If extensive simulations and test drives are conducted, then the thoroughness of verification is improved, but the cost and time requirements increase
Solution Approach 1:
The patent performs preliminary action by training artificial neural networks with extensive driving data before actual testing. This pre-training phase allows the system to learn critical patterns and edge cases, enabling more efficient subsequent testing with fewer simulations and test drives needed to achieve thorough verification.
Solution Approach 2:
The patent introduces artificial neural networks as an intermediary between raw driving data and test case generation. This intermediary processes and filters information to identify the most critical test scenarios, reducing the need for extensive simulations and test drives while maintaining verification thoroughness.
Data Source
AI summary
The invention relates to a computer-implemented method for approximating a subset of test results of a virtual test of a device for the at least partially autonomous guidance of a motor vehicle. The invention further relates to a test unit for approximating a subset of test results of a virtual test of a device for the at least partially autonomous guidance of a motor vehicle. The invention also relates to a computer program and a computer-readable data carrier.