AV Stack Validation Using QoRE-Aware Driving Scenario Generation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current autonomous vehicle (AV) technologies face challenges in validating their AI stacks due to limited training data, especially in complex and dangerous scenarios, and lack high-fidelity photorealistic simulation data, which restricts their ability to perform well in varied environmental conditions.
Innovation Solution
A system and method that utilize a Quality of Ride Experience (QoRE)-aware cognitive engine to generate and simulate driving scenarios based on Operational Design Domain (ODD) and real-world data, allowing for iterative validation of Advanced Driver Assistance Systems (ADAS) and AVs, with adjustable complexity levels to assess performance metrics.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If simulation testing is used to validate AV stack, then testing coverage and scalability are improved, but availability of high-fidelity photorealistic data is limited
Solution Approach 1:
The patent uses synthetic data generation to create photorealistic simulation data that copies the characteristics of real-world data. The system generates high-fidelity photorealistic images, point clouds, and sensor data through rendering engines and simulation environments, eliminating the need for expensive and time-consuming real-world data collection while maintaining data quality for validation purposes
Solution Approach 2:
The patent introduces an intermediary validation framework that bridges simulation and real-world testing. The system uses synthetic data as an intermediary to pre-validate AV algorithms before deployment, creating a intermediate testing layer that reduces reliance on both pure simulation and expensive real-world testing while improving overall validation efficiency
2Reliability
If real-world testing is used to validate AV stack, then data authenticity is improved, but cost and setup time increase
Solution Approach 1:
The patent applies preliminary action by performing validation using synthetic data before real-world deployment. The system pre-trains and validates AV perception, planning, and control algorithms using generated photorealistic data, ensuring algorithms are ready before actual road testing. This preliminary validation reduces the time and cost of real-world testing while maintaining data authenticity through the use of realistic synthetic scenarios
3Ease of operation
If conventional testing methods are used, then implementation simplicity is maintained, but ability to capture environmental variations and complexities is limited
Solution Approach 1:
The patent implements dynamics by creating a flexible simulation framework that can dynamically adjust environmental parameters, scenario complexity, and data generation settings. The system allows users to modify scene configurations, weather conditions, traffic patterns, and sensor characteristics without reimplementation, maintaining ease of operation while dramatically increasing adaptability to various environmental variations and complex scenarios
Data Source
AI summary
This disclosure relates to method and system for validating an Autonomous Vehicle (AV) stack. The method may include receiving an Operational Design Domain (ODD) and real-world data for evaluating at least one of an Advanced Driver Assistance System (ADAS) and the AV. The ODD is based on at least one feature of at least one of the ADAS and the AV. For each of a plurality of iterations, the method may further include generating a driving scenario based on the ODD of the AV and the real-world data through a Quality of Ride Experience (QoRE)-aware cognitive engine, plugging and running at least one of the ADAS and the AV algorithm based on the driving scenario, and determining a set of performance metrics corresponding to the at least one feature of at least one of the ADAS and the AV in the driving scenario based on the simulating.