A Comprehensive Examination of Naturalistic Lane-Changes

Abstract

Michael Goodman was the NHTSA Task Order Manager for this project. 16. Abstract This research effort provided valuable insight into the nature and severity of lane changes in a naturalistic driving environment. Sixteen commuters who normally drove more than 25 miles (40 km) in each direction participated. The two research vehicles were a sedan and an SUV; each participant drove each vehicle for ten days. Data gathering was automatic, and no experimenter was present in the vehicle. There were 8,667 lane changes observed over 23,949 miles of driving, making this the largest known data collection effort for the study of lane changes. Analysis of the full data set resulted in many interesting findings regarding the frequency, duration, urgency, and severity of lane changes in regard to maneuver type, direction, and other classification variables. A subset of the full data set (500 lane changes) was then analyzed in greater depth using the sensor data collected by the instrumented vehicle. The sampled lane changes were generally of the more severe and urgent types since these are the cases in which a lane change collision avoidance system is likely to be of greatest help.Variables analyzed for the sampled lane changes included turn signal use, braking behavior, steering behavior, eye glance patterns, and forward and rearward area analysis. The concept of a safety envelope for lane changes was then developed using the forward and rearward area analyses. Finally, the data were used to provide recommendations for designers of lane change CAS in terms of display location and activation criteria. Overall, the research described in this report provides insight into the behaviors and parameters associated with lane changes, while the naturalistic data archive has the potential to address other questions related to driving behavior.17.

On the Roles of Eye Gaze and Head Dynamics in Predicting Driver's Intent to Change Lanes

“…The NHTSA Lane Change Study [3] and the Intelligent Vehicle Highway System review [41] both led to the hypothesis that measuring head dynamics may be enough to detect distinctive behavioral cues prior to a lane change. Having confirmed that hypothesis, it can be argued that the relative ease of capturing head motion information, compared with eye gaze in vehicles, outweighs the advantages of adding eye data to head, lane, and vehicle data.…”

“…Correspondingly, during right lane changes, drivers looked at the right mirror with 36%-53% probability and the rearview mirror with 82%-92% probability. Moreover, the mirror glance duration before lane change maneuvers lasted, on average, 1.1 s, varying between 0.8 and 1.6 s [3]. Mourant and Donohue observed that lengthy blind spot checks occurred only in conjunction with lane-change maneuvers; in lane-keeping situations, no such checks were performed by the drivers [14].…”

“…The participants in the study were told about the goals of the study, however, potentially changing their behavior. In fact, the participants' near-ubiquitous usage of turn signals to indicate lane changes was significantly higher than turn indicator usage from naturalistic studies of drivers' lane-change behaviors [3]. As described here, the data used in our study come from experiments in which drivers were not told about the goals of the study, increasing the likelihood of naturalistic behaviors.…”

“…Simulators though do not capture all the dynamics and variability of real-world environments [4]. Some real-world studies of driver behavior during lane changes have measured eye gaze by manually reducing data [3], [4], [15], [23], [24]. By doing so, they can ensure the reliability and accuracy of the eye-gaze data; these studies have shown some promise of using eye gaze as a cue for driver intent.…”

“…Therefore, to approximate the ideal case and retain the best possible chance of getting reliable and accurate eyegaze estimates, the data were manually reduced. The procedure followed was similar to those followed in the National Highway Traffic Safety Administration (NHTSA) lane-change and workload experiments [3], [4] to produce output that a real-world eye-gaze tracker would output in an optimal setting.…”