Road Model Data Coverage Planning for AV Simulation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Generating detailed road model data for autonomous vehicle testing is computationally expensive and resource-intensive, making it inefficient to cover all test scenarios, especially when underlying data changes frequently or requires high fidelity.
Innovation Solution
An optimizer is used to select the least number of map sections to generate, ensuring a desired coverage amount for the test suite by identifying pre-existing data and determining the optimal locations for new data generation, thereby reducing resource usage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If detailed road model data is generated for all test scenarios, then simulation accuracy is improved, but computational cost and resource usage increase significantly
Solution Approach 1:
The system applies different data generation strategies to different map sections based on their specific needs. High-fidelity road model data is generated only for map sections that are frequently traversed or critical for test scenarios, while lower-fidelity or pre-existing data is used for less critical areas. This localized approach maintains simulation accuracy where needed while reducing overall computational cost.
Solution Approach 2:
Instead of generating complete detailed road model data for all possible test scenarios, the system generates data for a subset of map sections that provides sufficient coverage for the desired simulation accuracy. The optimizer identifies the minimum necessary set of map sections to generate, accepting that not all scenarios will have perfect data coverage, thus reducing computational resources while maintaining adequate simulation quality.
2Productivity
If road model data is generated for frequent test scenarios, then test coverage is improved, but resource usage increases when underlying data changes frequently
Solution Approach 1:
The system pre-generates road model data for map sections that are likely to be needed based on historical test scenario frequencies and requirements. By anticipating which map sections will be needed and preparing their data in advance, the system improves test coverage readiness while avoiding the costly repeated generation that would occur if data were created on-demand for each test scenario.
Solution Approach 2:
The system dynamically adjusts the fidelity and detail level of generated road model data based on the specific requirements of different test scenarios. For scenarios requiring high precision, more detailed data is generated, while for less demanding scenarios, lower-fidelity data suffices. This parameter adjustment allows the system to maintain good test coverage while optimizing resource consumption by matching data quality to scenario needs.
Data Source
AI summary
Autonomous vehicles are tested and evaluated on the road and in simulated road environments. When testing and evaluating autonomous vehicles in a simulated road environment, having detailed three-dimensional geometries and characteristics of road surfaces in the form of road model data can greatly improve a simulated environment's ability to test and evaluate vehicle controls and dynamics, and can generate a more accurate pose for the autonomous vehicle in simulation. However, generating road model data is computationally expensive, and it is not desirable to generate road model data for an entire map to cover all test scenarios. An optimizer can reduce the amount of resources needed by selecting a least number of map sections to be generated and determining the locations of the map sections to be generated that achieves a desired coverage amount for the test suite.