Crowdsourced Virtual Driving Scenarios for AD/ADAS Corner Cases
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for testing autonomous driving (AD) and advanced driver assistance systems (ADAS) require extensive real-world testing, which is time-consuming and costly, and struggle to expose vehicles to rare corner case scenarios effectively.
Innovation Solution
A server-based method simulates a virtual environment with mixed human and computer-controlled virtual objects, allowing users to generate scenarios that test AD/ADAS functionality, leveraging crowdsourcing to increase the variety and frequency of corner case scenarios.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If extensive real-world testing is conducted to validate AD/ADAS functionality, then reliability of the system is improved, but loss of time and productivity deteriorate significantly
Solution Approach 1:
The patent creates virtual copies of real-world driving scenarios, corner cases, and traffic environments in a simulated world. Instead of physically testing billions of kilometers in the real world, the system generates and tests scenario data copies in virtual environments, maintaining validation reliability while dramatically reducing time loss.
Solution Approach 2:
The system performs preliminary identification and generation of corner case scenarios and test data before actual testing begins. By pre-generating diverse scenario data including rare corner cases in the virtual world, the system prepares comprehensive test cases in advance, eliminating the need for prolonged real-world searching and testing.
2Loss of time
If aggressive testing assumptions are made to reduce testing time, then loss of time is reduced, but measurement precision of corner case scenarios deteriorates
Solution Approach 1:
The system dynamically adjusts the generation and selection of scenario data to ensure comprehensive coverage of corner cases. Rather than using fixed aggressive testing assumptions, the virtual environment adaptively generates and prioritizes rare corner case scenarios based on predefined criteria, maintaining measurement precision while reducing overall testing time through intelligent dynamic scenario selection.
3Reliability
If more vehicles are deployed for real-world testing to increase corner case exposure, then corner case identification is improved, but device complexity and cost increase
Solution Approach 1:
The patent creates a universal virtual testing platform that can evaluate multiple AD/ADAS systems simultaneously using the same infrastructure. Instead of requiring separate real-world testing fleets for each vehicle system, the simulated world serves as a multi-functional testing environment where diverse scenario data is generated once and reused across multiple validation efforts, eliminating the need for extensive additional physical vehicles.
4Productivity
If virtual simulation is used to reduce testing costs, then productivity is improved, but difficulty of detecting and measuring real-world scenarios increases
Solution Approach 1:
The system changes the parameters of scenario generation and selection to ensure virtual scenarios accurately reflect real-world conditions. By adjusting scenario parameters, diversity criteria, and validation metrics in the virtual environment to match real-world distributions and characteristics, the system maintains scenario realism and detectability while achieving cost and time efficiency through virtual simulation.
Data Source
Figure 1
Figure 2a~2b
Figure 3
AI summary
Method and server for supporting generation of scenarios for testing autonomous driving and/or advanced driver assistance system, AD/ADAS, functionality for real world vehicles. A server (101; 500) provides (301) a virtual environment (200) simulating an environment relevant for operation of vehicles having said AD/ADAS functionality and in which is operating: fully computer controlled movable virtual objects (230a-c), human controlled movable virtual objects (220a-c) and at least one virtual AD/ADAS vehicle (210) operating according to said AD/ADAS functionality. The server (101; 500) allows devices (101-103) to remotely connect to the server (105; 500) and users of said devices (101-103) to, via user interfaces of the devices (101-103), control said human controlled movable virtual objects (220a-c), respectively, in the virtual environment (200), and thereby cause generation of scenarios that said at least one virtual AD/ADAS vehicle (210) is subjected to.