Ye Mengqiu, Osman Osama A, Ishak Sherif, Hashemi Bita
Department of Civil and Environmental Engineering, Louisiana State University, Patrick F. Taylor Hall, Baton Rouge, LA 70803, United States.
Department of Civil and Environmental Engineering, Louisiana State University, Patrick F. Taylor Hall, Baton Rouge, LA 70803, United States.
Accid Anal Prev. 2017 Sep;106:385-391. doi: 10.1016/j.aap.2017.07.010. Epub 2017 Jul 15.
Distracted driving has long been acknowledged as one of the leading causes of death or injury in roadway crashes. The focus of past research has been mainly on the impact of different causes of distraction on driving behavior. However, only a few studies attempted to address how some driving behavior attributes could be linked to the cause of distraction. In essence, this study takes advantage of the rich SHRP 2 Naturalistic Driving Study (NDS) database to develop a model for detecting the likelihood of a driver's involvement in secondary tasks from distinctive attributes of driving behavior. Five performance attributes, namely speed, longitudinal acceleration, lateral acceleration, yaw rate, and throttle position were used to describe the driving behavior. A model was developed for each of three selected secondary tasks: calling, texting, and passenger interaction. The models were developed using a supervised feed-forward Artificial Neural Network (ANN) architecture to account for the effect of inherent nonlinearity in the relationships between driving behavior and secondary tasks. The results show that the developed ANN models were able to detect the drivers' involvement in calling, texting, and passenger interaction with an overall accuracy of 99.5%, 98.1%, and 99.8%, respectively. These results show that the selected driving performance attributes were effective in detecting the associated secondary tasks with driving behavior. The results are very promising and the developed models could potentially be applied in crash investigations to resolve legal disputes in traffic accidents.
长期以来，分心驾驶一直被认为是道路交通事故中导致死亡或受伤的主要原因之一。过去的研究重点主要在于不同分心原因对驾驶行为的影响。然而，只有少数研究试图探讨某些驾驶行为属性如何与分心原因相关联。本质上，本研究利用丰富的SHRP 2自然istic驾驶研究（NDS）数据库，开发了一个模型，用于根据驾驶行为的独特属性检测驾驶员参与次要任务的可能性。使用速度、纵向加速度、横向加速度、横摆率和油门位置这五个性能属性来描述驾驶行为。针对三个选定的次要任务（打电话、发短信和与乘客互动）分别开发了一个模型。这些模型是使用有监督的前馈人工神经网络（ANN）架构开发的，以考虑驾驶行为与次要任务之间关系中固有的非线性影响。结果表明，所开发的人工神经网络模型能够分别以99.5%、98.1%和99.8%的总体准确率检测驾驶员参与打电话、发短信和与乘客互动的情况。这些结果表明，所选的驾驶性能属性在检测与驾驶行为相关的次要任务方面是有效的。结果非常有前景，所开发的模型有可能应用于事故调查，以解决交通事故中的法律纠纷。
