Using Perception of Alignment and Short-Term Memory to Understand Middle-aged Driver's Driving Speed on Mountainous Highway

Despite numbers of researches on driving speed, there is limited work to study it from the aspect of driver's visual information. The paper presents a new methodology to model vehicle speed on four-lane divided mountainous highway. The targeted drivers are middle-aged adults. Continuous speed data and road videos were recorded by tachograph during naturalistic driving experiments, from which a large amount of data of visual lane parameters were extracted. In the speed prediction model, new input variables are defined as the weighted average values of visual lane parameters of the preceding road and local road. The new variables were proven to be useful explanatory variables for the proposed model. Based on the regression result, driver's Short-Term Memory can be described as concave weighting distribution. Data analysis showed that middle-aged drivers had consistent Short-Term Memory behaviors. The process of Short-Term Memory use preceding information ranging from 28s to 34s, during which the majority of useful information maintains within the latest 15s. Importantly, the model is built directly from driver's visual information, instead of curvature or gradient. The findings provide insight into how driver views and processes road geometric information. This improvement could contribute to future studies of road design in accordance with driver's perception and expectation.