一句话：提出开放领域特定建模语言 SML4ADS，以非编程、用户友好的方式表示并生成可执行自动驾驶测试场景。
自动驾驶系统上市前需进行充分安全评估，但仅靠道路测试难以覆盖足够里程，因此场景测试成为关键，而灵活生成场景仍是挑战。本文提出面向自动驾驶的领域特定建模语言 SML4ADS，用于场景表示与生成。与已有工作相比，该方法以非编程、用户友好的方式简化场景描述，支持车辆随机行为建模，并能生成可在 CARLA 中执行的可执行场景。作者将 SML4ADS 应用于多种典型场景，初步验证了其在场景建模与可执行场景生成方面的有效性与可行性。
Autonomous Driving Systems(ADS) require extensive evaluation of safety before they can come onto the market. However, since relying solely on field testing is practically infeasible due to the impossibility to cover sufficient distances to ensure adequate safety, the focus shifted to scenario-based testing. The challenge is to generate scenarios flexibly. We proposed Scenario Modeling Language for ADS (SML4ADS) as a Domain-Specific Modeling Language (DSML) for scenario representation and generation. Compared to other existing works, our approach simplifies the description of scenarios in a non-programming, user-friendly manner, allows modeling stochastic behavior of vehicles and generating executable scenario in CARLA. We apply SML4ADS in numerous typical scenarios to preliminarily demonstrate the effectiveness and feasibility of our approach in modeling and generating executable scenarios.
@inproceedings{paperbot7237,
title = {SML4ADS: An Open DSML for Autonomous Driving Scenario Representation and Generation},
author = {Bo Li and Dehui Du and Sicong Chen and Minjun Wei and Chenghang Zheng and Xinyuan Zhang},
booktitle = {ASE},
year = {2022},
doi = {10.1145/3551349.3561169}
}