Autonomous Driving Test Case Replay for Reliable Dynamic Verification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional dynamic verification methods for autonomous driving systems lack reliability due to the use of automatically generated test cases that fail to accurately reflect real driving environments, leading to unreliable results in ensuring the safety and functionality of autonomous driving software.
Innovation Solution
A method and apparatus for generating a reliable test case for dynamic verification of autonomous driving systems, which involves receiving input signals from sensors, re-sampling and rearranging them, comparing output signals to determine confirmed test cases, applying weights, and transmitting these to an external device for real-time verification, ensuring the test cases accurately reflect real driving conditions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If conventional automatic test case generation tools are used for dynamic verification, then the verification process can be automated, but the reliability of verification results deteriorates because test cases fail to accurately reflect real driving environments
Solution Approach 1:
The patent creates a virtual copy of the real driving environment by recording actual sensor data, vehicle states, and control commands during real driving, then replaying this recorded data as test cases. This copying approach preserves the authenticity of real driving scenarios while enabling automated verification without requiring actual vehicles on the road.
Solution Approach 2:
The system performs preliminary data collection and processing during actual driving operations, storing sensor data, vehicle states, and control commands in advance. These pre-collected data are then reused as test cases for subsequent verification processes, eliminating the need to recreate driving scenarios and ensuring consistency with real conditions.
2Reliability
If manually created test cases are used to accurately reflect real driving environments, then the reliability of verification results improves, but the complexity and time consumption of test case generation increases
Solution Approach 1:
The system automatically captures and processes its own operational data during real driving. Sensors, vehicle controllers, and communication buses self-generate the test case data without external intervention. This self-service approach eliminates manual data collection while preserving the authenticity of real driving scenarios.
Solution Approach 2:
The system continuously monitors and records the feedback loop between control commands issued by the autonomous driving controller and the actual vehicle responses. This feedback mechanism captures the complete closed-loop behavior of the system, ensuring test cases reflect real operational dynamics without requiring manual analysis or simplification.
3Measurement precision
If comprehensive sensor data and vehicle state information are collected for test case generation, then the accuracy and reliability of verification improve, but the data processing complexity and computational load increase
Solution Approach 1:
The system pre-processes and structures sensor data, vehicle states, and control commands during the data collection phase, organizing them into standardized formats suitable for replay. This preliminary organization reduces the computational burden during test case execution and verification, as data is already prepared and indexed for efficient access.
Solution Approach 2:
Instead of processing raw sensor data in real-time during verification, the system creates compact copies of the complete driving scenario including all sensor readings, vehicle states, and control commands. These copied datasets preserve measurement precision while reducing processing complexity during test execution by eliminating the need for real-time data acquisition and synchronization.
Data Source
AI summary
A method and an apparatus for generating a test case (TC) for dynamic verification of an autonomous driving system are provided. The method includes receiving an input signal and generating a temporary TC based on the input signal. A confirmed TC is determined based on a first output signal corresponding to the input signal and a second output signal corresponding to the temporary TC. A final TC is determined by re-arranging the confirmed TC and the final TC is transmitted to an external device.