Macroscopic relaxation after on-ramps in real data and in cellular automata simulations

This paper presents a methodology for the testing of lane changing rules in traffic simulation models. The idea is to use simple macroscopic data to test microscopic rules. More precisely, the relaxation of the lane flow rates after an on-ramp on a multilane highway is taken as an integrated signature of the lane-changing efficiency.Our approach is two-fold. First the characteristic length scale on which multilane traffic relaxes after a perturbation (here an on-ramp) is measured under various conditions on real data. It is shown to be of the order of a few hundred meters. In the free flow regime, this relaxation length was found to increase with the occupancy when the perturbation due to the on-ramp is weak, while it becomes independent from the occupancy for strong on-ramp perturbations. By contrast, in the congested regime, the relaxation length was found to decrease with increasing occupancy.The comparison of experimental observations with simulation results allows to test indirectly the relevance of the lane changing rules used in the model. Here we apply this approach to cellular automata based simulations. We evidence and discuss several shortcomings of the numerical method.

► Methodology proposal: use the relaxation of lane flow rates after an on-ramp as a test for lane-changing rules.
► In the free flow regime, relaxation length is found to increase with occupancy for a weak on-ramp perturbation, and to be constant for a strong on-ramp perturbation.
► In the congested regime, relaxation length is found to decrease with occupancy.
► The application of the methodology to cellular automata simulations shows in particular the importance of having mixed vehicle types.