Vehicle Control Software Verification Using Parallel Output Comparison
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing autonomous driving systems face challenges in achieving comprehensive verification of control software due to the complexity of operation conditions and the reliance on pre-designed test cases, particularly when machine learning is involved, leading to potential precision issues in varying situations.
Innovation Solution
A vehicle control device that executes old and new versions of control software sequentially or in parallel, using idle resources to compare output values and identify execution conditions affecting precision, allowing for comprehensive verification without hindering real-time operations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If pre-designed test cases are used for verification, then verification process is simple, but verification comprehensiveness is insufficient
Solution Approach 1:
The patent creates virtual copies of the actual vehicle operating environment by recording real sensor data, GPS information, and control outputs during vehicle operation. These recorded data sets are then used as test inputs for verifying control software in different scenarios, enabling comprehensive verification without manually designing test cases for each scenario
Solution Approach 2:
The patent performs preliminary data collection during actual vehicle operation by recording sensor inputs, GPS positions, and control software outputs. This pre-collected data is stored and later used for verification purposes, allowing the system to prepare verification materials in advance without needing to design test cases beforehand
2Measurement precision
If control software is optimized for specific situations, then precision for that situation improves, but precision in other situations deteriorates
Solution Approach 1:
The patent creates a universal verification system that can evaluate control software performance across multiple different situations using the same verification framework. By collecting diverse operating data during vehicle operation and using it for verification, the system can assess whether software optimizations for specific situations adversely affect performance in other situations
3Reliability
If comprehensive test cases covering all operation conditions are designed, then verification comprehensiveness improves, but test case design complexity increases
Solution Approach 1:
The verification system automatically generates test cases from recorded vehicle operation data without requiring manual intervention to design test scenarios. The system self-services by collecting real operating data, processing it into verification inputs, and automatically executing verification tests, thereby achieving comprehensive coverage without complex manual test case design
Data Source
AI summary
A vehicle control device includes a verification management unit that executes old control software unit 112 representing an old version of control software and new control software unit 113 representing a new version of control software in sequence or in parallel, and an output verification unit that when an output value from old control software unit 112 and an output value from new control software unit 113 do not match, outputs information indicating the output values' not matching.