Integrity of laser-based feature extraction and data association
Integrity of laser-based feature extraction and data association
复制标题
基于激光的特征提取和数据关联的完整性
DOI：
--
复制
发表时间：
2016
期刊：
影响因子：
--
通讯作者：
B. Pervan
中科院分区：
文献类型：
--
作者：
M. Joerger;M. Jamoom;M. Spenko;B. Pervan
在本文中，一个新的完整性风险评估方法的开发和测试的激光和雷达为基础的导航算法，使用特征提取（FE）和数据关联（DA）。这项工作的目的是安全关键的自主车辆导航。FE和DA是两个预估计器测量处理步骤，旨在重复和一致地识别环境中的地标。FE和DA的主要安全风险是由不正确的关联（将一个地标误认为另一个地标）引起的。为了评估这种风险，首先在FE引入了一个标准：它建立了标志之间的最小归一化间隔，确保它们可以可靠地、量化地区分。然后，一个创新为基础的DA过程的设计，它提供了一种手段来评估不正确的关联的概率，同时考虑所有潜在的测量排列。这些算法进行了分析和测试，显示不正确的关联对安全风险的影响。
In this paper, a new integrity risk evaluation method is developed and tested for laser and radar-based navigation algorithms using feature extraction (FE) and data association (DA). This work is intended for safety-critical autonomous vehicle navigation. FE and DA are two pre-estimator measurement processing steps that aim at repeatedly and consistently identifying landmarks in the environment. A major risk for safety in FE and DA is caused by incorrect associations (mistaking one landmark for another). To assess this risk, a criterion is first introduced at FE: it establishes the minimum normalized separation between landmarks ensuring that they can be reliably, quantifiably distinguished. Then, an innovation-based DA process is designed, which provides the means to evaluate the probability of incorrect associations while considering all potential measurement permutations. These algorithms are analyzed and tested, showing the impact of incorrect associations on safety risk.