Driving Scenario Probability Metrics for Autonomous Vehicle Safety
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing planning systems for autonomous and semi-autonomous vehicles face challenges in efficiently and accurately testing their performance and safety in diverse driving scenarios, particularly due to the time-consuming and resource-intensive process of updating scenario data and probability metrics, which often leads to reduced accuracy as map data and vehicle control systems change.
Innovation Solution
Techniques for dynamically and automatically generating scenario metrics and data using abstracted representations of vehicle maneuvers and map data, allowing for efficient computation and regular updates of safety metrics by querying databases of driving events, without the need for manual input or extensive raw data searches.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual updates of scenario data and probability metrics are performed, then accuracy of safety metrics is maintained, but time consumption and resource intensity increase significantly
Solution Approach 1:
The system automatically queries databases of driving events and map data to generate updated scenario metrics without requiring manual intervention. The planning system self-updates by retrieving relevant data, computing probability metrics, and refreshing scenario definitions autonomously, thereby maintaining accuracy while eliminating manual labor and associated time costs.
Solution Approach 2:
The system pre-queries databases for driving events and map data before simulation runs are needed. By proactively retrieving and processing scenario data in advance, the system prepares updated metrics ahead of time, reducing the time required during actual safety testing and deployment cycles.
2Reliability
If comprehensive scenario data is collected from raw data searches, then completeness of driving scenarios is improved, but computational resources are excessively consumed
Solution Approach 1:
The system extracts only the specific driving events and map data that are relevant to the current simulation scenario from the comprehensive databases. By querying for scenario-specific data rather than processing all available raw data, the system achieves complete and accurate scenario coverage while minimizing computational resource consumption through targeted data extraction.
3Measurement precision
If frequent updates of scenario metrics are performed, then accuracy reflects current driving conditions, but computational overhead increases
Solution Approach 1:
The system performs updates of scenario metrics at periodic intervals or triggered by specific events such as map data changes or simulation milestones. This periodic update approach ensures that scenario metrics remain accurate and reflective of current driving conditions while avoiding unnecessary computational overhead from continuous or overly frequent updates.
Data Source
AI summary
Techniques for determining a metrics associated with a vehicle are disclosed. The techniques may comprise receiving a driving simulation scenario associated with a simulated vehicle. Based at least in part on the scenario and a dataset of recorded driving associated with operation of a test vehicle, a plurality of driving events present in the dataset may be determined. Based at least in part on the plurality of driving events, a first metric associated with a likelihood of occurrence of the driving simulation scenario may be determined. A simulation may be instantiated based at least in part on the driving simulation scenario. Based at least in part on the simulation, a safety metric indicative of the safe performance of the simulated vehicle in the simulation may be determined. An aggregate safety metric may be determined based at least in part on the safety metric and the first metric.