Autonomous Vehicle Simulation with Dynamic Environmental Entity Generation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current autonomous driving simulation systems rely on limited manual editing or data from road tests, restricting the effectiveness and diversity of simulation test resources due to human limitations in creating scenarios and behaviors for NPCs in autonomous vehicle simulations.
Innovation Solution
A virtual platform simulation method that automatically generates simulation parameters for environmental entities, such as update periods and numbers, within a preset area, enabling the creation of diverse and dynamic traffic scenarios to test autonomous driving systems, including the ability to restart and interact with main entities.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If manual editing or road test data is used to define NPC behaviors, then the simulation system can be implemented, but the diversity and effectiveness of simulation test resources are limited
Solution Approach 1:
The simulation system automatically generates NPC behavior data through self-service mechanisms. The system uses the autonomous vehicle's own test data and simulation results to train models that generate new NPC behaviors, eliminating the need for extensive manual editing while continuously expanding test resource diversity
Solution Approach 2:
The system changes parameters by using different training configurations, model architectures, and data processing methods to generate diverse NPC behaviors from the same base data. By adjusting parameters such as behavior patterns, response times, and interaction rules, the system creates varied simulation scenarios without increasing system complexity
2Productivity
If more simulation test data is generated to improve test coverage, then the effectiveness of autonomous vehicle testing increases, but the time and computational resources required increase
Solution Approach 1:
The system performs preliminary actions by pre-training models with available data and pre-generating NPC behavior patterns before actual simulation tests. This allows rapid generation of diverse test scenarios during execution, improving test coverage efficiency without proportionally increasing data generation time
Solution Approach 2:
The system maintains continuity by running simulations and data generation processes continuously. Test data from one simulation immediately feeds into the next, creating an uninterrupted workflow that maximizes computational resource utilization and reduces idle time, thereby improving overall productivity
Data Source
AI summary
The present disclosure relates to a simulation method, a computing device, and a storage medium for use in automatic generation of environmental entities around a target test object in a simulation platform, thereby improving the simulation test efficiency. The simulation method includes: generating a main entity including a representation of an autonomous vehicle in a simulation platform; acquiring simulation parameters of environmental entities, the simulation parameters including update periods of the environmental entities and a number constraint of the environmental entities in a preset area within each update period; determining, according to the number constraint, an expected number of the environmental entities in the preset area within each update period; and generating, according to the simulation parameters and the expected number, a corresponding number of the environmental entities in the preset area within each update period so as to enable the newly-generated environmental entities to run in the simulation platform.