01624362

Component

Driving performance deteriorates with excess driving stress rises, which may increase vehicle accident likelihood. This study aims to quantify the effect of driving stress by monitoring the heart rate increase with various traffic conditions in a real-world road network. The data collection includes electrocardiogram, vehicle GPS trajectories, road conditions from video, and vehicle conditions from CAN bus. The authors propose a machine learning methodology based on Random Forest for the estimation of car driver stress due to different driving events. In contrast to other statistical methods and machine learning methods, Random Forest can handle different types of predictor variables, make a high accurate prediction and give variable importance analysis. Results indicate that average speed, coefficient of covariance of speed, frequency of brake operation and frequency of acceleration operation contribute about 78% relative importance to driving stress. Further sensitivity analysis show that low average speed, large speed variance, frequent operations of brake and acceleration will cause high level of driving stress. Based on the proposed model, a driving heat map is drawn in a large-scale road network, which can be applied to a safety-based route guidance system.

This paper was sponsored by TRB committee AND30 Standing Committee on Simulation and Measurement of Vehicle and Operator Performance.

01618707

17-02363

English

17p

2017

Transportation Research Board 96th Annual Meeting

Digital/other

Figures; Maps; Photos; References (36) ; Tables

Automatic data collection systems; Driver performance; Driver support systems; Highway safety; Risk assessment; Route guidance; Stress (Physiology); Stress (Psychology); Vehicle trajectories

Highways; Safety and Human Factors

Transportation Research Board Annual Meeting 2017 Paper #17-02363

TRIS, TRB, ATRI

Dec 8 2016 10:53AM