Empirics and Models of Fragmented Lane Changes

Existing microscopic traffic models represent the lane-changing maneuver as a continuous and uninterrupted lateral movement of the vehicle from its original to the target lane. We term this representation as Continuous Lane-Changing (CLC). Recent empirical studies find that not all lane-changing maneuvers are continuous; the lane-changer may pause its lateral movement during the maneuver resulting in a Fragmented Lane-Changing (FLC). We analysed a set of 1064 lane changes from NGSIM dataset which contains 270 FLCs. In comparison to a CLC, this study investigates the distinction of an FLC in terms of its execution and its effects on neighbouring vehicles. We find that during the execution of an FLC, the lane-changer exhibits distinct kinematics and takes a longer duration to complete the lane-changing. We propose a trajectory model to describe the lateral kinematics during an FLC. Additionally, we find that the FLC induces a distinct effect on the follower in the target lane, and propose a model to describe the transient behavior of the target-follower during an FLC. The modelling results suggest that the accuracy of traffic flow models can be improved by deploying lane change execution and impact models that are specific to FLC and CLC. Besides, this study identifies a set of factors that might be related to the decision-making process behind FLC: an average driver executes an FLC when the preceding and following vehicles in the target lane are slower, and when the follower in the target lane is closer than those observed during the onset of a CLC. Our findings suggest that FLC is motivated by an increased necessity to change lane such as during a mandatory lane change.