ABSTRACT

Although attention to data fusion has undergone rapid growth since the late 1980s, there are still relatively few applications in transportation management. Most research has based fusion weight estimation on the variance of each data source, assigning high weights to low variance data, implying that low variance means high accuracy. We propose a data fusion methodology where weights are assigned in a way data variance and sensor bias are minimized, but also consistency among data sources is maximized. The proposed methodology is flexible to work with multiple data sources, with different reliability and variability, and under different traffic conditions. The inclusion of consistency is shown to be statistically significant during special events and incidents and demonstrates its validity in successfully representing changes in traffic patterns by reasonably estimating their magnitude. Results from a case study that validate this method are shown.

摘要

尽管自 80 年代后期以来对数据融合的关注经历了快速增长，但在交通管理中的应用仍然相对较少。大多数研究基于每个数据源的方差进行融合权重估计，为低方差数据分配高权重，这意味着低方差意味着高准确性。我们提出了一种数据融合方法，其中以最小化数据方差和传感器偏差的方式分配权重，同时最大化数据源之间的一致性。所提出的方法可以灵活地处理具有不同可靠性和可变性以及不同交通条件的多个数据源。在特殊事件和事故中，一致性的包含在统计上是显着的，并通过合理估计它们的大小来证明其在成功表示交通模式变化方面的有效性。显示了验证此方法的案例研究结果。