Autonomous Driving Scenario Mining from Abnormal Vehicle Trajectories
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for testing autonomous vehicle (AV) safety and performance are inadequate, as they require extensive real-world driving data to identify and simulate realistic, unusual driving scenarios, making it impractical to achieve human-level safety standards.
Innovation Solution
A method for analyzing driving behavior data to identify abnormal trajectories, creating a normal driving behavior model, and using object tracking to extract and simulate realistic driving scenarios, allowing for systematic and scalable mining of challenging scenarios without relying on actual test miles.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If real-world driving data is collected to identify unusual driving scenarios, then the realism and representativeness of training scenarios is improved, but the cost and time required for data collection increases significantly
Solution Approach 1:
The patent creates synthetic copies of driving scenarios by simulating abnormal trajectories and embedding them in realistic driving contexts. Instead of collecting extensive real-world data, the system generates artificial but realistic scenarios that replicate the essential characteristics of unusual driving situations, thereby reducing data collection time while maintaining scenario realism.
Solution Approach 2:
The system performs preliminary analysis of normal driving patterns to establish baseline behavior models before generating abnormal scenarios. By pre-processing normal trajectory data to identify typical driving patterns, the system can then efficiently synthesize unusual scenarios that deviate from these patterns, avoiding the need for extensive real-world data collection of rare events.
2Reliability
If the number of test miles is increased to find sufficient instances of unsafe behaviour, then the safety testing thoroughness is improved, but the cost and time required increases exponentially
Solution Approach 1:
The patent creates synthetic copies of dangerous driving scenarios by generating abnormal trajectories and embedding them in realistic driving contexts. This allows comprehensive safety testing of rare but critical failure modes without requiring millions of actual test miles, thereby maintaining testing thoroughness while dramatically improving testing efficiency.
Solution Approach 2:
The system performs preliminary identification of abnormal driving patterns using unsupervised learning on normal trajectory data. By pre-characterizing what constitutes abnormal behavior, the system can efficiently generate and test against relevant safety scenarios without exhaustively searching through vast amounts of real-world test data.
3Adaptability or versatility
If unsupervised learning is used to identify abnormal trajectories from normal driving data, then the ability to discover challenging scenarios is improved, but the complexity of data processing increases
Solution Approach 1:
The patent introduces normal driving behavior models as intermediaries between raw trajectory data and abnormal scenario identification. These models serve as mediators that capture typical driving patterns, allowing the system to efficiently identify deviations without directly analyzing all raw data, thereby reducing processing complexity while maintaining versatile scenario discovery.
Solution Approach 2:
The system segments the data processing task into distinct stages: first learning normal patterns from trajectory data, then identifying deviations from these patterns. This segmentation allows each stage to focus on specific aspects of the problem, reducing overall processing complexity while maintaining the ability to discover diverse challenging scenarios.
Data Source
AI summary
One aspect herein provides a method of analysing driving behaviour in a data processing computer system, the method comprising: receiving at the data processing computer system driving behaviour data to be analysed, wherein the driving behaviour data records vehicle movements within a monitored driving area; analysing the driving behaviour data to determine a normal driving behaviour model for the monitored driving area; using object tracking to determine driving trajectories of vehicles driving in the monitored driving area; comparing the driving trajectories with the normal driving behaviour model to identify at least one abnormal driving trajectory; and extracting a portion of the driving behaviour data corresponding to a time interval associated with the abnormal driving trajectory.