Modeling driver's braking and steering behavior along horizontal curves of two-lane rural highways for ADAS...

Drivers continuously manipulate the vehicle's steering and pedals (throttle, brake) while driving. Inadequate or excessive manipulation leads to crashes, possibly leading to injuries. An Advanced Driving Assistance System (ADAS) can warn drivers of inadequate or excessive manipulation to avert risky situations. Such systems evolve continuously and need more road infrastructure inputs for better driving assistance. Rural highways showcased more fatalities than urban counterparts, with more road departure crashes along horizontal curves than straight sections. Hence, the present study explores drivers' braking and steering behavior along the horizontal curves of two-lane rural highways and proposes a suitable runoff risk model.

Sixty-eight drivers participated in a driving simulator study, where they drove along fifty-two horizontal curves in free-flow conditions. Typical five brake and three steering measures found in past studies were evaluated. Based on data distribution of brake and steering measures, Tobit and mixed-effects multiple linear regression models were developed to understand the association of these measures with the approach speed and geometric parameters of horizontal curves. Further, a generalized linear mixed runoff (road departure) risk model with a logit link function was developed to identify the effective measures among all brake and steering variables.

The brake and steering measures were significantly associated with the approach speed and the geometric parameters such as curve radius, gradient, and turn type (left or right). Further, the runoff risk analysis revealed a significant effect of braking distance (BD) and standard deviation of steering wheel position (SDSWP). The interaction between BD and SDSWP was significant, suggesting that the runoff risk increased with SDSWP; however, it decreased with BD.

The runoff risk intensified with higher steering instability along horizontal curves, while it can be reduced with effective brake pedal usage. The study suggests better usage of brakes than steering before entering the curve. It potentially reduces the vehicle speed, thus ensuring more time to perceive the curve and help reduce the driver's steering instability. The models developed in this study can be used in the ADAS systems upon validation with the field observations.