Autonomous Driving Test Course With AI Obstacle Scenario Verification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional traveling test courses are inadequate for testing autonomous driving vehicles, particularly when using AI, as they do not account for various obstacles and conditions that a human driver can easily navigate, leading to insufficient verification of the autonomous driving system's performance.
Innovation Solution
A test method and program that involve setting multiple obstacles on a test course, using sensors and AI to acquire information, and performing a traveling test to verify the autonomous driving vehicle's behavior and control, enabling precise control at the nanosecond level through deep learning and Level 6 calculation power.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional traveling test courses are used for autonomous driving vehicles, then the test course structure is simple and easy to set up, but the verification of autonomous driving system performance is insufficient
Solution Approach 1:
The patent applies preliminary action by pre-setting multiple types of obstacles (static obstacles like walls and cones, dynamic obstacles like moving vehicles and pedestrians) on the test course before conducting autonomous driving tests. This allows the autonomous driving system to be verified in advance against various realistic scenarios including obstacle detection, avoidance, and navigation, thereby improving verification reliability without requiring complex real-time test course modifications
Solution Approach 2:
The test course is segmented into multiple functional zones with different obstacle configurations and test scenarios. Each zone focuses on specific autonomous driving capabilities (e.g., obstacle detection zone, avoidance maneuver zone, navigation zone), allowing systematic verification of different system aspects while maintaining manageable course complexity
2Extent of automation
If AI-based autonomous driving control is implemented, then the autonomous driving capability is enhanced, but the control precision requirement increases to nanosecond level
Solution Approach 1:
The patent replaces traditional mechanical control systems with AI-based autonomous driving control that processes sensor data and makes driving decisions through machine learning models. This substitution enables enhanced autonomous capability by using deep learning algorithms to interpret complex environments and generate control commands, with timing precision achieved through software-based nanosecond-level control signal generation rather than mechanical timing mechanisms
Solution Approach 2:
The system implements continuous feedback loops where sensors constantly monitor the environment and vehicle state, AI algorithms process this information in real-time, and control commands are adjusted accordingly. This closed-loop feedback mechanism maintains nanosecond-level control precision by continuously correcting deviations and adapting to changing conditions, ensuring reliable autonomous driving performance
3Adaptability or versatility
If multiple obstacles are set on the test course to simulate real-world conditions, then the test realism is improved, but the test course complexity and setup time increase
Solution Approach 1:
The patent employs dynamic obstacles that can move autonomously or be controlled remotely, allowing the same physical obstacle to appear in multiple positions and configurations. This dynamic capability enables diverse test scenarios (moving vehicles, pedestrians, varying obstacle patterns) using the same set of physical objects, thereby improving test realism without proportionally increasing setup time or course complexity
Solution Approach 2:
The system uses virtual copies or digital representations of obstacles in conjunction with physical obstacles. Virtual obstacles can be programmed to appear, disappear, or move without physical manipulation, allowing rapid scenario changes and realistic test conditions while minimizing the time and effort required for physical course setup and reconfiguration
Data Source
Figure 1
Figure 2
Figure 3
AI summary
An autonomous driving test method includes setting a plurality of obstacles on a test course, acquiring information regarding the obstacles, and performing a traveling test related to autonomous driving of an autonomous driving vehicle by using a plurality of pieces of the acquired information and AI.