Autonomous Driving Scenario Testing With Dynamic Obstacle Resource Pools
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current autonomous driving simulation software faces challenges in building complex scenarios, including low verification efficiency, inability to accurately respond to real road emergencies, and managing diverse obstacles, which hinders the development of reliable autonomous driving systems.
Innovation Solution
An obstacle resource pool is introduced to efficiently and centrally manage obstacles, allowing for the creation of complex test scenarios by storing and retrieving motion information of various obstacles, including refined track, navigation-type, macro traffic flow, and road test obstacles, enhancing simulation authenticity and interaction realism.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If real-vehicle road tests are used for algorithm verification, then the reliability of autonomous driving system is improved, but the research and development cost increases and algorithm development efficiency decreases
Solution Approach 1:
The patent creates virtual copies of real road scenarios, obstacles, and traffic conditions through simulation software. By copying real-world data into virtual environments, the system maintains high reliability for algorithm verification while eliminating the need for costly and time-consuming real-vehicle road tests, thus improving algorithm development efficiency
Solution Approach 2:
The patent performs preliminary actions by pre-building comprehensive obstacle libraries and scenario databases before actual testing. This preparation includes collecting real road data, creating virtual obstacle models, and establishing test scenarios in advance, allowing efficient algorithm verification without repeated real-world testing
2Ease of manufacture
If simulation software is used to build single simulation scenarios, then the research and development cost is reduced, but the verification efficiency of complex scenarios decreases and real road conditions cannot be accurately restored
Solution Approach 1:
The patent merges multiple single simulation scenarios into complex composite scenarios by combining various obstacle types, traffic conditions, and environmental factors. The obstacle library integrates diverse obstacle models (pedestrians, vehicles, animals, static objects) that can be combined to create realistic complex scenarios, improving verification efficiency while maintaining cost-effectiveness
Solution Approach 2:
The patent applies parameter changes by adjusting obstacle motion parameters, environmental conditions, and scenario configurations to restore real road conditions. By modifying parameters such as obstacle speed, trajectory, density, and interaction patterns, the simulation accurately reproduces complex real-world scenarios without requiring expensive real-vehicle testing
3Device complexity
If multiple types of obstacles are managed separately, then the management complexity is reduced, but the ability to simulate dynamic interaction and complex scenarios decreases
Solution Approach 1:
The patent creates a universal obstacle library that serves multiple functions: storing individual obstacle data, managing complex scenario compositions, and enabling dynamic interactions. This unified obstacle management system handles diverse obstacle types (movable, immovable, dynamic, static) within a single framework, reducing management complexity while enhancing simulation versatility and interaction capabilities
Data Source
AI summary
The method includes: obtaining N obstacles from an obstacle set based on a test function of the autonomous-driving algorithm, and storing an ID and initial motion information of each of the N obstacles to an obstacle resource pool; obtaining motion information of a surrounding obstacle of a first obstacle from the obstacle resource pool based on an ID of the first obstacle, and controlling moving of the first obstacle based on the motion information of the surrounding obstacle of the first obstacle, to obtain motion information of the first obstacle; storing the motion information to the obstacle resource pool; obtaining a result of interaction between an ego vehicle and n obstacles from the obstacle resource pool based on location information of the ego vehicle; and determining a test result of the autonomous-driving algorithm based on the interaction result.