Analysis of Driver's Steering Behavior for Lane Change Prediction

With the development of intelligent transportation systems and advanced driver assistance systems (ADAS), early prediction of driver's intentions is required. Among that, lane change is a complex driver conducted maneuver, thus, improper execution could lead to severe traffic accidents. Therefore, it is urgent to develop lane change prediction systems, especially which could work in the initial phase of lane change. Many previous studies have developed lane change detection algorithms, however, they use different criterions as onset of lane change: (1) the onset of lane change maneuver, (2) the time that vehicle crosses the lane line. This makes it hard to compare the earliest time where these systems could detect lane change reliably. Therefore, a systematically comparison of the predicting time in these algorithms is required, as well, parameters which could be used for early lane change detection need to be highlighted. In the present study, a driving simulator study of lane change was conducted, with 18 drivers involved. Data of vehicle's lateral position, steering wheel angle and velocity during both lane change and lane keep phase were collected. The results showed that, the time gap between the two criterions is approximately 3 s, thus, the earliest predicting time of these algorithms ranged from the onset of lane change maneuver to the first peak of steering wheel angle. In addition, a further analysis of drivers' steering behavior of lane changing showed that, means and variances of steering wheel velocity, in the phase up to 0.55 s before the onset of lane change maneuver, were significantly different from that in lane keep phase. This suggests steering wheel velocity a potential parameter for early lane change detection.