Drivers-in-the-Loop Simulation Platform for Autonomous Vehicle Testing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for testing autonomous driving systems, such as traditional closed field tests and open road tests, are inadequate for achieving the required large test mileage and are economically and time-consuming.
Innovation Solution
A multi-drivers-in-the-loop driving testing platform that integrates a sensing simulation system, a vehicle dynamic simulation system, a driving simulator, and a scene simulation system to generate object-level perception information, simulate vehicle dynamics, provide a driving environment, and update the driving scene in real time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional closed field tests and open road tests are used for autonomous vehicle testing, then the testing can be conducted with real vehicles, but the test mileage requirement of hundreds of millions to billions of miles cannot be achieved and the economic cost and time cost become unbearable
Solution Approach 1:
The patent uses virtual sensors to generate object-level perception information and creates virtual driving scenes that replicate real-world driving conditions. This copying approach allows the system to achieve billions of miles of testing virtually, eliminating the time and resource constraints of physical road tests while maintaining testing effectiveness.
Solution Approach 2:
The system pre-generates comprehensive driving scene libraries and perception information databases before actual testing begins. By preparing virtual testing environments and data sets in advance, the system can rapidly conduct extensive mileage testing without the setup and execution time required for traditional road tests.
2Productivity
If traditional closed field tests and open road tests are used for autonomous vehicle testing, then the testing can be conducted with real vehicles, but the economic cost becomes unbearable
Solution Approach 1:
By creating virtual copies of driving scenes, vehicles, and sensor data, the system eliminates the need for expensive physical test tracks, test vehicles, and extensive human driver resources. The virtual environment replication provides cost-effective alternative to millions of dollars in traditional testing infrastructure.
Solution Approach 2:
The patent replaces the mechanical physical testing system with a computational virtual testing system. Instead of physically moving vehicles through test tracks and relying on human drivers, the system uses computer-generated perception information and virtual scene rendering to achieve the same testing objectives at fraction of the cost.
3Productivity
If virtual simulation testing is used for autonomous driving system development, then the test mileage requirements can be achieved, but relevant research and development capabilities are lacking at present
Solution Approach 1:
The patent divides the virtual simulation system into distinct functional modules: virtual sensor systems for generating perception information, vehicle control systems for autonomous driving algorithms, and scene simulation systems for creating driving environments. This modular segmentation allows the platform to adapt to different testing scenarios and be enhanced with new capabilities as technology develops.
Solution Approach 2:
The virtual simulation platform is designed to perform multiple functions: it can simulate various sensor types (cameras, radar, lidar), support different autonomous driving control algorithms, generate diverse driving scenes, and accommodate both perception algorithm testing and control algorithm validation, making it a versatile research and development tool.
Data Source
AI summary
A multi-drivers-in-the-loop driving testing platform is provided including at least a sensing simulation system, a vehicle dynamic simulation system, a driving simulator and a scene simulation system. The sensing simulation system is configured to generate object-level perception information, and send it to a vehicle control system; the driving simulator is configured to provide a driving environment and a driving scene for a human driver, output a driving instruction according to a driving intention of the human driver, and then send the driving instruction to the vehicle control system; the vehicle dynamic simulation system is configured to calculate state information according to control signals output by the vehicle control system; and the scene simulation system is configured to update the driving scene displayed in the driving simulator timely according to the vehicle state information. The disclosure saves costs of research and development and shortens cycles of research and development.