Scenario Simulation From Recorded Vehicle Data for AV Validation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for testing and validating autonomous vehicle controllers are inefficient, as they require manual enumeration of numerous scenarios, which is time-consuming and may not cover all relevant conditions, potentially leading to untested scenarios and limited insight into the vehicle's operational space under varying conditions.
Innovation Solution
The use of previously recorded sensor data to generate high-quality simulation scenarios, allowing for the creation of realistic simulated environments that mimic real-world conditions, including sensor data, perception data, and prediction data, to test and validate autonomous vehicle controllers.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If manual enumeration of test scenarios is used, then scenario coverage can be controlled, but the testing process becomes time-consuming and inefficient
Solution Approach 1:
The system performs preliminary actions by automatically generating comprehensive test scenarios from historical sensor data before actual testing begins. This pre-generation of scenarios using real-world recorded data eliminates the need for time-consuming manual enumeration while ensuring comprehensive coverage of edge cases and rare events that may have been missed in manual testing.
Solution Approach 2:
The system creates copies of real-world scenarios by reconstructing simulated environments from recorded sensor data. These simulated scenarios are copies of actual driving conditions, allowing efficient replay and analysis without requiring physical recreation of each scenario, thus dramatically improving testing productivity.
2Reliability
If manual scenario enumeration is used, then testing can be performed, but comprehensive coverage of operational space is limited
Solution Approach 1:
The system performs self-service by automatically generating, organizing, and managing test scenarios without human intervention. The automated system extracts scenarios from historical data, simulates them, and analyzes results, eliminating the complexity of manual scenario enumeration while achieving comprehensive coverage of the operational space through systematic processing of all available data.
Solution Approach 2:
The system changes parameters by varying environmental conditions, vehicle states, and object behaviors in simulated scenarios based on historical data distributions. This systematic parameter variation ensures comprehensive coverage of operational space across multiple dimensions while automating the process, thereby improving reliability without increasing manual complexity.
3Reliability
If more test scenarios are generated, then validation coverage improves, but computational resources increase
Solution Approach 1:
The system applies partial action by selectively generating and prioritizing test scenarios based on their importance and rarity. Instead of exhaustively simulating every possible scenario, the system focuses computational resources on high-value scenarios identified from historical data, achieving effective validation coverage with optimized resource utilization.
Solution Approach 2:
The system uses efficient copying techniques by creating simplified simulated representations of complex real-world scenarios. These simulated copies retain essential characteristics needed for validation while requiring fewer computational resources than full-scale physical testing or highly detailed simulations, enabling broader scenario coverage with constrained resources.
Data Source
AI summary
A vehicle can capture data for use in a simulator. Objects represented in the vehicle data can be instantiated in simulation and move according to object models/controllers. A user can tune how closely simulated motion of the object corresponds to the previously recorded data based on simulation costs. The scenarios can be used for testing and validating interactions and responses of a vehicle controller within a simulated environment. The scenarios can include simulated objects that traverse the simulated environment and perform actions based on the captured data and/or interactions within the simulated environment. Objects observed by the vehicle can be disregarded from simulation based on one or more filters. A user can override, augment, or otherwise modify simulations instantiated based on the one or more filters and captured data.