This paper discusses the feasibility of data captured in a long-term Naturalistic Driving Study (NDS) for identification of vehicle dynamics. Driving data were captured for over a year. In this data capture, there was minimal effort to define or control everyday driving practices. While the use of real-world data for model parameter identification is a well-known method, NDS are commonly used to explore the behavior of drivers or to analyze real-world traffic situations. Data from NDS have not yet been used for the purpose of parameterizing vehicle dynamics models since everyday drives commonly do not reflect the full range of vehicle dynamics. This leads to the question if the data from an NDS contains the needed information to describe vehicle dynamics accurately. This paper shows that data captured from long-term everyday vehicle usage is sufficient to characterize vehicle dynamics models. It uses lateral vehicle dynamics as an example to show how the data quantity changes the model accuracy and robustness. There is a point where any further data capture produces redundancy and does not add to the overall information. The well-known single-track model serves as the modeling example which offers options to simply compare the derived model behavior with a reference.

中文翻译：

---

使用自然驾驶研究来识别车辆动力学

本文讨论了在长期自然驾驶研究 (NDS) 中捕获的数据用于识别车辆动力学的可行性。驾驶数据被捕获了一年多。在此数据捕获中，定义或控制日常驾驶实践的工作量很小。虽然使用真实世界数据进行模型参数识别是一种众所周知的方法，但 NDS 通常用于探索驾驶员的行为或分析真实世界的交通情况。来自 NDS 的数据尚未用于对车辆动力学模型进行参数化，因为日常驾驶通常不能反映车辆动力学的全部范围。这就引出了一个问题，即来自 NDS 的数据是否包含准确描述车辆动力学所需的信息。本文表明，从长期日常车辆使用中捕获的数据足以表征车辆动力学模型。以横向车辆动力学为例，展示了数据量如何改变模型的准确性和鲁棒性。在某一点上，任何进一步的数据捕获都会产生冗余，并且不会增加整体信息。著名的单轨模型作为建模示例，它提供了将派生模型行为与参考进行简单比较的选项。