Modeling the Effects of Warning Lead Time, Warning Reliability and Warning Style on Human Performance under Connected Vehicle Settings
Yiqi Zhang,Changxu Wu
DOI: https://doi.org/10.1177/1541931218621158
2018-01-01
Proceedings of the Human Factors and Ergonomics Society Annual Meeting
Abstract:Deaths and injuries resulted from traffic accidents is still a major public health problem. Recent advances in connected vehicle technology support a connected driving environment in which vehicles are enabled to communicate with each other and with roadside infrastructures via Dedicated Short Range Communication (DSRC). Connected vehicle safety applications supported by this technology allow drivers to learn about the traffic situations out of their sight and ahead of time so that drivers are warned early enough to make proper responses. As the connected vehicle systems (CVS) are designed with an aim to improving driver safety, the effectiveness of the CVS can not be achieved without drivers making proper responses in responding to the wireless warnings. Therefore, it is essential to understand and model the mechanism for human processing and responding to warnings from connected vehicle systems, and apply the driver model to optimize the design the CVS at the interface level and the communication level. Queuing Network-Model Human Processor (QN-MHP) is a computational framework that integrates three discrete serial stages of human information processing (i.e., perceptual, cognitive, and motor processing) into three continuous subnetworks. Each subnetwork is constructed of multiple servers and links among these servers. Each individual server is an abstraction of a brain area with specific functions, and links among servers represent neural pathways among functional brain areas. The neurological processing of stimuli is illustrated in the transformation of entities passing through routes in QN-MHP. Since this architecture was established, QN-MHP has been applied to quantify various aspects of aspects of driver behavior and performance, including speed control (Bi & Liu, 2009; Zhao & Wu, 2013b), lateral control (Bi et al., 2012; Bi et al., 2013), driver distraction (Bi et al., 2012; Fuller, Reed & Liu, 2012; Liu, Feyen & Tsimhoni, 2006), and driver workload (Wu & Liu, 2007; Wu et al., 2008). Most of the driver model built upon QN-MHP focused on the modeling of driver performance in normal driving situation. In a previous work of authors, a mathematical model was developed to predict the effects of warning loudness, word choice, and lead time on drivers’ warning reaction time (Zhang, Wu, & Wan, 2016). The current research focused on the development of a mathematical model based on QN-MHP to quantify and predict driver performance in responding to warnings from connected vehicle systems, including warning response time and the selection of warning response type. The model also quantified the effects of important warning characteristics in connected vehicle systems, including warning reliability, warning lead time, and speech warning style. The model was validated via an experimental study indicating its good predictability of driver behavior and performance in connected vehicle systems. In particular, the model was able to explain 68.83% of the warning response type in the initial trial of the experiment with a root mean square error (RMSE) of 0.18. By adding the warning effect on the probability of a response type through trials, the model was able to explain 65.13% of the warning response type in the initial trial of the experiment with a root mean square error (RMSE) of 0.16. In terms of warning response time, the model prediction of warning response time under different warning reliability, style and lead time were very similar to the response time results from the experiments. The model was able to explain 88.30% of the experimental response time in average with a root mean square error (RMSE) of 0.16s. The developed driver model could be applied to optimize the design of the connected vehicle systems based on driver