In this simulation of 500 m of multi-lane highway, the red cars are fast (have a average desired speed of 25 m/s) and the blue cars are slow (have an average desired speed of 15 m/s). A variation of Gipps Car Following model is used. For the lane changing, a forward search algorithm with 8-second time horizon is used.
TRB 2007
PhD Dissertation
Tactical Lane Change Model With Sequential Maneuver Planning. N.A. Webster, T. Suzuki, M. Kuwahara. Transportmetrica, Volume 4, Issue 1 2008, pp. 63-78
Vehicles build a tree of possible outcomes based on their lane change actions now and until the time horizon, and pick the current action which gets them the farthest ahead, with preference for returning to the travel (left) lane. The lines shown are each vehicle's current best path, based on the distance gained as well as being in the shoulder (left) lane at the end of the time horizon.
Note: If the cars do not move at first in your browser, try reloading (refreshing) the page.