Detecting intermittent steering activity: Development of a phase-detection algorithm

Drivers usually maintain an error-neglecting control strategy (passive phase) in keeping their vehicle on the road, only to change to an error-correcting approach (active phase) when the vehicle state becomes inadequate. We developed an algorithm that is capable of detecting whether the driver is currently error-neglecting or error-correcting in straight lane keeping tasks. The development of this algorithm was part of a larger research project, DrivObs, that aims at creating an advanced driver observation tool. Performance of the algorithm in a straight lane driving task with lateral vehicle position perturbations was tested in a Monte Carlo simulation using Matlab/Simulink. Results show that the algorithm is capable of correctly detecting active or passive phase 90-95% of the time, depending on vehicle speed and algorithm settings.