Development of kinks in car-following models

TL;DR

This paper analyzes the stability and wave solutions of car-following traffic models near the absolute stability threshold, revealing the limitations of traditional solvability analysis and proposing a multiple-time-scales approach.

Contribution

It demonstrates that the standard solvability analysis does not uniquely select kink solutions and introduces a multiple-time-scales method to better understand traffic wave dynamics.

Findings

Only one kink solution is preserved by corrections, indicating a form of selection.

A two-parameter family of traveling wave solutions includes kinks as a special case.

Multiple-time-scales analysis reveals conditions for decay, evolution, or pairing of traffic inclusions.

Abstract

Many car-following models of traffic flow admit the possibility of absolute stability, a situation in which uniform traffic flow at any spacing is linearly stable. Near the threshold of absolute stability, these models can often be reduced to a modified Korteweg-deVries (mKdV) equation plus small corrections. The hyperbolic-tangent "kink" solutions of the mKdV equation are usually of particular interest, as they represent transition zones between regions of different traffic spacings. Solvability analysis then shows that only a single member of the one-parameter family of kink solutions is preserved by the correction terms, and this is interpreted as a kind of selection. We point out that one cannot extend the solvability analysis to a multiple-time-scales calculation, so that the solvability analysis does not point the way to any dynamical mechanism by which the "selected" kink might actually be selected. On the other hand, we display a two-parameter family of traveling wave solutions of the mKdV equation which describe regions of one traffic spacing embedded in traffic of a different spacing; this family includes the kink solutions as a limiting case. We carry out the multiple-time-scales calculation and find conditions under which the inclusions decay, conditions that lead to a selected inclusion, and conditions for which the inclusion evolves into a pair of kinks. Finally, we show that the usual "solvability" calculation for kink solutions does not in fact identify a selected kink, but -- for all kink solutions -- merely gives a first-order correction to the relation between the traffic spacings far behind and far ahead of the kink.