David LeBlanc

A WARNING AND INTERVENTION SYSTEM TO PREVENT ROAD-DEPARTURE ACCIDENTS

SUMMARY A driver assistance system is described which provides an automatic road-departure warning and intervention function. The system is intended for drivers who are drifting off the roadway at a shallow angle of departure due to inattention, drowsiness, intoxication, or other causes. An integrated set of model-based estimation algorithms serves to anticipate imminent road departures. Interventions are achieved using differential braking -- the application of unequal brake torques at the right and left wheels. The driver remains in the loop, always able to provide the primary steering commands. A driver state assessment module is being developed to further adapt to individual driven and to identify changes in driver behavior. We describe the algorithms, present simulation results, and include preliminary experimental data from a prototype vehicle.

Lane Geometry Perception and the Characterization of Its Associated Uncertainty

This paper addresses the reconstruction of down-range road geometry from imaging sensors for application to motor vehicle active safety systems. This study assumes measurements of lane marker locations in the previewed scene are available from an maging sensor. An algorithm is developed to extend the perception ra of a single-far-field sensor to alleviate the field of view problem Two steady-state Kalman filters and a least square curve fitting scheme are developed to compute estimates of the road geometry. Simulations are used to compare the performance of the different road modeling schemes for different roadway scenarios, providing insights useful for selecting model-based road geometry estimation techniques Finally an algorithm to characterize the uncertainty in road geometry perception is proposed and validated.

Spectral Power Analysis of Drivers’ Gas Pedal Control during Steady-state Car-following on Freeways:

This paper investigated the frequency characteristics of drivers’ gas pedal control in steady-state car-following on freeways by using vehicle sensor data from an existing naturalistic driving study. The main objectives were to examine the frequency range and distributions of a driver operating the gas pedal when following a lead vehicle, and whether the higher and lower frequency components of the gas pedal signal would vary when following a lead vehicle with varying distances. A total of 1,461 driving segments each with 90-seconds of steady-state freeway car-following were extracted from the naturalistic driving data. Fourier analysis was performed to convert the time series data of drivers’ gas pedal control to the frequency domain. The results show that during steady-state freeway car-following, the power of the gas pedal control peaks at around 0.033 Hz or 15 s per pedal movement (derived using the median of the peak frequency), and the upper limit of the frequency is around 0.94 Hz or 0.5 s per pedal movement (derived using the 95th percentile of the cutoff frequency). Further analysis showed that following a lead vehicle with smaller gap was associated with a larger proportion of the higher frequency component (p < .001), and following a lead vehicle with larger gap was associated with a larger proportion of the lower frequency component (p < .001). This suggests that the larger gap may allow the driver to relax control of the gas pedal with smoother operation. Potential applications of this paper include developing more realistic driver models that could be used in designing advanced driver assistance systems.