Probabilistic Lane Changing Models based on Logistic Functions

This paper presents a logistic function-based framework for modeling lane-changing behavior on multi-lane freeways. The mandatory lane change (MLC) model computes the lane-changing probability as a logistic function of the remaining distance to the destination, with an inflection point that scales linearly with the number of required lane changes, generalizing naturally to freeways of any width. The discretionary lane change (DLC) model applies the same logistic structure to the speed differential between adjacent lanes. Both models differentiate human-driven vehicle (HDV) and autonomous vehicle (AV) behavior through parameter selection alone. A closing-speed gap acceptance criterion completes the framework with speed-dependent safety requirements. A rate rule converts the logistic value into a probability over the exposure accumulated between evaluations (distance driven for MLC, time elapsed for DLC), making lane-change frequency invariant to the evaluation rate; without it, a 10 Hz automated vehicle makes 78% more mandatory changes. The models are integrated into an event-driven cellular automata microsimulation and evaluated against NGSIM trajectory data, two established lane-changing models (MOBIL, Gipps), and mixed-traffic scenarios. At default parameters the baselines produce higher throughput (6365 and 6242 vs. 6041 veh/h); sensitivity analysis shows that the mandatory inflection point governs this gap.