.png){kind=icon}
返回
Distributed Accelerated Concurrency Testing Method for Automated Driving System
.png){kind=icon}
PRE
AI
DOI：10.1109/TVT.2023.3295366.png)
.png){kind=icon}
原文求助.png){kind=icon}
分享.png){kind=icon}
收藏
摘要
En 中文
Recently, the scenario-based test method has become essential for the safety verification of automated vehicles (AVs). However, with the increased functionality of automated driving systems (ADSs) and the expansion of the operational designed domain (ODD), the number of scenario elements that need to be tested has rapidly increased, leading to an explosion of concrete scenarios to be tested. Accordingly, the accelerated test methods targeting the enhanced generation of hazardous concrete scenarios have become the main trend of AV tests. However, existing methods mainly adopt serial processing, which cannot provide multi-threaded parallelism. This paper proposes a distributed accelerated concurrency testing (ACT) method for ADS, which combines an accelerated test method with traditional concurrent testing. The test frame is divided into three layers: a top-layer manager, middle-layer scheduler, and bottom-layer executor. Then, the specific tasks of roles at different layers are determined; for instance, the coverage of a middle-layer scheduler and the test parameters of a bottom-layer executor. In addition, the modified Ackley function is used for preliminary verification of the effectiveness of the proposed test method. Finally, a black-box AV algorithm is tested in the front vehicle cut-in scenario in a virtual environment using the proposed ACT method, traversal concurrency testing method, accelerated serial testing method, and traversal serial testing method. The test results show that the proposed ACT method can effectively find dangerous concrete scenarios in the logical scenario parameter space and reduce the testing time significantly compared to the other three methods.
Keyword:
Automated vehicle
scenario-based test
accelerated test
parallelism test
particle swarm algorithm
期刊
IF:
7.1
论文数:
1.8W
被引数:
6.6W
机构
引用论文
The femtosecond-to-second photochemistry of red-shifted fast-closing anion channelrhodopsin PsACR1
OAAI
OAAI分享收藏Testing Scenario Library Generation for Connected and Automated Vehicles, Part I: Methodology联网和自动化车辆的测试场景库生成，第一部分: 方法OAAI
OAAI分享收藏Coexistence of Lambert‐Eaton myasthenic syndrome and subacute cerebellar degeneration.png)
Neurology
NeurologyIF0
PREAI
分享收藏Testing Scenario Library Generation for Connected and Automated Vehicles: An Adaptive Framework联网和自动驾驶车辆的测试场景库生成: 自适应框架OAAI
OAAI分享收藏Parallel Computing and Multicore Platform to Assess Electric Vehicle Hosting Capacity并行计算和多核平台评估电动汽车托管能力
PREAI
分享收藏
查看更多内容