Risk Quantification for Automated Driving using Information from V2V Basic Safety Messages
Risk Quantification for Automated Driving using Information from V2V Basic Safety Messages
复制标题
DOI：
10.1109/vtc2021-spring51267.2021.9448849
复制
发表时间：
2021-04
期刊：
影响因子：
--
通讯作者：
Raghvendra V. Cowlagi;Rebecca C. Debski;A. Wyglinski
中科院分区：
文献类型：
--
作者：
Raghvendra V. Cowlagi;Rebecca C. Debski;A. Wyglinski
使用来自V2 V链路的数据沿着车载传感器数据被认为是迈向未来自动驾驶安全性和可靠性的关键一步。我们解决的自我车辆的短期轨迹规划。我们提出了一个威胁场模型的交通周围的自我车辆。非正式地，威胁字段指示车辆附近发生碰撞的可能性。我们研究的不确定性，由于周围车辆的位置和速度的不确定性，这可能是已知的自我CAV通过基本的安全信息的威胁领域。轨迹的成本由预期的威胁暴露来定义，并且轨迹的风险基于成本的方差来量化。不确定性量化研究使用蒙特卡罗抽样以及扰动为基础的方法。本文的主要结果是观察到，小的定位误差和/或速度测量误差可能会导致计划轨迹的大风险。
Using data from V2V links along with onboard sensor data is recognized as a crucial step towards the safety and reliability of future automated driving. We address short-term trajectory planning for the ego vehicle. We propose a threat field model of the traffic surrounding the ego vehicle. Informally, the threat field indicates the possibility of collisions in the vehicle’s vicinity. We study uncertainty in the threat field due to uncertainty in the positions and velocities of surrounding vehicles, which may be known to the ego CAV via basic safety messages. The cost of trajectories is defined by the expected threat exposure, and the risk of trajectories is quantified based on the variance in cost. Uncertainty quantification is studied using Monte Carlo sampling as well a perturbation-based approach. The main result of this paper is the observation that small localization errors and/or speed measurement errors can lead to large risks in planned trajectories.