Critical Discussion of "Synchronized Flow", Simulation of Pedestrian Evacuation, and Optimization of Production Processes
Extracted source text ยท arXiv cond-mat/0210097
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% This is a sample input file for your contribution to a
% multi-author book ("proceedings-like) to be published
% by Springer Verlag.
%
% Please use it as a template for your own input, and please
% follow the instructions for the formal editing of your
% manuscript as described in the file "1readme".
%
% Please send the Tex and figure files of your manuscript
% together with any additional style files as well as the
% PS file to the editor of your book.
%
% He or she will collect all contributions for the planned
% book, possibly compile them all in one go and pass the
% complete set of manuscripts on to Springer.
%
% Vers. 07/99
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%RECOMMENDED%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\documentclass[runningheads]{svmult}
%\usepackage{makeidx} % allows index generation
\usepackage{graphicx} % standard LaTeX graphics tool
% for including eps-figure files
%\usepackage{subeqnar} % subnumbers individual equations
% within an array
%\usepackage{multicol} % used for the two-column index
%\usepackage{cropmark} % cropmarks for pages without
% pagenumbers - only needed when manuscript
% is printed from paper and not from data
%\usepackage{physprbb} % modified textarea for proceedings,
% lecture notes, and the like.
%\makeindex % used for the subject index
% please use the style sprmidx.sty with
% your makeindex program
%%upright Greek letters (example below: upright "mu")
\newcommand{\greeksym}[1]{{\usefont{U}{psy}{m}{n}#1}}
\newcommand{\umu}{\mbox{\greeksym{m}}}
\newcommand{\udelta}{\mbox{\greeksym{d}}}
\newcommand{\uDelta}{\mbox{\greeksym{D}}}
\newcommand{\uPi}{\mbox{\greeksym{P}}}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%OPTIONAL%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
%\usepackage{amstex} % useful for coding complex math
%\mathindent\parindent % needed in case "Amstex" is used
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%AUTHOR_STYLES_AND_DEFINITIONS%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
%Please reduce your own definitions and macros to an absolute
%minimum since otherwise the editor will find it rather
%strenuous to compile all individual contributions to a
%single book file
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\newcommand{\tit}[1]{} %{`#1',}
\newcommand{\disstit}[1]{`#1',}
%\newcommand{\tit}[1]{}
%\newcommand{\disstit}[1]{}
\newcommand{\pag}[2]{pp.~#1}
\newcommand{\art}[5]{{\em #1} {\bf #2}, #3 #5}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\newcommand{\bib}[1]{\bibitem{#1}}
\begin{document}
\unitlength=1.0cm
%
\title*{Critical Discussion of ``Synchronized Flow'',\protect\newline
Simulation of Pedestrian Evacuation, and\protect\newline
Optimization of Production Processes}
%
%
\toctitle{Critical Discussion of ``Synchronized Flow'', Simulation of\protect\newline
Pedestrian Evacuation, and Optimization of Production Processes}
% allows explicit linebreak for the table of content
%
%
\titlerunning{``Synchronized Flow'', Pedestrian Evacuation, and Optimal Production}
% allows abbreviation of title, if the full title is too long
% to fit in the running head
%
\author{Dirk Helbing\inst{1}
\and Ill\'{e}s J. Farkas\inst{2}
\and Dominique Fasold\inst{1}
\and Martin Treiber\inst{1}
\and Tam\'{a}s Vicsek\inst{2}}
%
\authorrunning{Dirk Helbing et al.}
% if there are more than two authors,
% please abbreviate author list for running head
%
%
\institute{Institute for Economics and Traffic, Dresden University of Technology,
Andreas-Schubert-Str. 23, D-01062 Dresden, Germany
\and
Department of Biological Physics, E\"otv\"os University, Budapest,
P\'azm\'any P\'eter S\'et\'any 1A, H-1117 Hungary}
\maketitle % typesets the title of the contribution
\begin{abstract}
We critically discuss the concept of ``synchronized flow'' from a
historical, empirical, and theoretical perspective. Problems related to the
measurement of vehicle data are highlighted, and questionable interpretations
are identified. Moreover, we propose a quantitative and consistent theory
of the empirical findings based on a phase diagram of congested traffic states, which is universal
for all conventional traffic models having the same instability diagram and a
fundamental diagram. New empirical and simulation data supporting this approach
are presented as well. We also give a short overview of the various phenomena
observed in panicking pedestrian crowds relevant from the point of
evacuation of buildings, ships, and stadia.
Some of these can be applied to the optimization of production processes,
e.g. the ``slower-is-faster effect''.
\end{abstract}
\section{Freeway Traffic: ``Synchronized flow'', ``Pinch Effect'', and Measurement Problems}
\subsection{What is new about ``synchronized flow?''}
Congested traffic has been investigated for many decades because of its complex phenomenology.
Therefore, Kerner and Rehborn have removed the data belonging to wide
moving jams (see MLC and TSG in Figs.~\ref{Fig9} and \ref{Fig10}) and found
found that the remaining data of congested traffic
data still displayed a wide and two-dimensional scattering \cite{KerRe96b},
see Fig.~\ref{Fig4}(c). By mistake (see Figs.~\ref{Fig1}, \ref{Fig4}(c), and Sec.~\ref{widescat}),
they concluded that all models assuming a fundamental diagram were wrong
and defined a new state called {\em ``synchronized flow''} (``synchronized'' because of the typical
synchronization among lanes in congested traffic, see Fig.~\ref{Fig2}(a),
and ``flow'' because of flowing in contrast to standing traffic in fully developed jams).
Since then, Kerner suggests to classify {\em three phases}:
(1) {\em free flow,}
(2) {\em ``synchronized flow''}, and
(3) {\em wide moving jams} (i.e. {\em moving localized clusters} whose width in longitudinal
direction is considerably higher than the width of the jam fronts). In some
applied empirical studies, however, Kerner {\em et al.} additionally
distinguish a fourth state of stop-and-go traffic \cite{FOTO,KerReAl00}.
\par
\begin{figure}[htbp]
\begin{center}
\includegraphics[width=6cm]{fig1a.eps}
\includegraphics[width=6cm]{fig1b.eps}
\end{center}
\caption[]{(a) Flow-density relation for narrow moving jams simulated with a microscopic
traffic model. (b) The aggregated (1-minute) data corresponding to the narrow moving
jams displayed in (a) show a wide scattering and erratic movement
in the flow-density plane. By mistake, this is used to characterize
``synchronized flow''.\label{Fig1}}
\end{figure}
{\em Free flow} is characterized by the average desired velocity $V_0$ and,
therefore, by a strong correlation and quasi-linear relation $Q \approx \rho V_0$
between the local flow $Q$ and the {\em local} density $\rho$ \cite{NeubSaScSc99}.
It is also well-known that {\em wide moving jams} propagate with constant form and
(phase) velocity $C\approx -13$km/h \cite{MikaKrYu69,KerRe96a,CasMa99}.
Kerner found that this propagation is not
affected by bottlenecks or the presence of ``synchronized flow''.
Moreover, he showed that the outflow $Q_{\rm out}$ from wide jams is a
self-organized traffic constant as well \cite{KerRe96a,Kerner1}.
In contrast to wide moving jams, the flow inside of {\em ``synchronized flow''} remains finite,
and its downstream front is normally
fixed at the location of some bottleneck, e.g. an on-ramp. Therefore, ``synchronized flow''
basically agrees with
previous observations of queued or congested traffic (see, e.g., Refs.
\cite{Per86,Bank90,Bank91a} and the references therein). %,AgyHa91}).
In his patent \cite{KerKirReh}, Kerner
applies the queuing theory himself, which goes back to
the fluid-dynamic traffic model by Lighthill and Whitham \cite{LighWh55}.
\par
The synchronization of the
average velocities among neighboring lanes has been already described by Koshi {\em et al.}
\cite{KosIwOh83} (but see also Refs. \cite{MikaKrYu69,EdiFo58,ForMuSi67}).
It is true on a {\em macroscopic} level for {\em all} forms of congested traffic
including wide moving jams. Simulations have shown that this is
a result of lane changes \cite{Lee3LeKi98},
while the assumed reduction in the lane changing rate \cite{KerRe97} occurs only after
the speeds in neighboring lanes have been successfully balanced \cite{ShvHe99}.
On a {\em microscopic} scale,
over-taking maneuvers continue to exist almost at all densities \cite{HelHu98}, see Figs.~\ref{Fig2}(b), (c).
Nevertheless, the probability of lane changes drops considerably with increasing density,
when most of the road is used up by the vehicles' safety headways \cite{HelHu98},
but not in the postulated $Z$-shaped way \cite{Kerner3}. Due to the reduced opportunities
for overtaking and the related bunching of vehicles, the velocity variance
goes down with increasing density as well \cite{KerRe97,Hel97a,Hel97c}.
\par
\begin{figure}[htbp]
\begin{center}
\begin{picture}(12,9.7)(-1.6,0)
\put(0.15,10){\includegraphics[width=5.3\unitlength,height=8.4\unitlength,
angle=-90]{fig2a.eps}}
\put(-0.4,4.5){\includegraphics[height=5\unitlength,angle=-90]{fig2b.eps}}
\put(3.6,4.5){\includegraphics[height=5\unitlength,angle=-90]{fig2c.eps}}
\end{picture}
\end{center}
\caption[]{(a) The difference of the average velocity in the left and the right lane
vanishes at densities above 30 vehicles per kilometer, corresponding to
a macroscopic synchronization of the speeds \cite{Hel97a}.
(b), (c) The difference in the empirically determined velocities of cars and trucks,
however, show that overtaking maneuvers continue to exist even at higher
densities \cite{HelHu98}.
\label{Fig2}}
\end{figure}
The transition between free and congested traffic is of hysteretic nature,
i.e. the inverse transition occurs at a lower density and a higher average
velocity. This has been known for a long time \cite{TreitMy74,Pay84,Hal187}.
Kerner specifies that the transition is usually caused by a localized
and short perturbation of traffic flow that starts downstream of the
on-ramp and propagates upstream with a velocity of about $-13$\,km/h.
As soon as the perturbation passes the on-ramp, it triggers the breakdown
which spreads upstream in the course of time. The congested state can then persist for
several hours \cite{KerRe97}.
\par
Moreover, Kerner and Rehborn distinguish three types of ``synchronized flow'' \cite{KerRe96b},
which may be short-lived:
%
(i) Stationary and homogeneous states where both the average
speed and the flow rate are stationary
(see, e.g., also Refs. \cite{Hal1Ag91,PerYaBr98,Wes98}. Later on,
we will these {\em ``homogeneous congested traffic''} (HCT) \cite{HelHeTr99}.\\
%
(ii) States where only the average vehicle speed is stationary,
named {\em ``homogeneous-in-speed states''} (see also
Refs. \cite{Ker98b,Lee3LeKi00}).
We interpret this state as {\em ``recovering traffic''} \cite{Review},
as it bears several signatures of free traffic and mostly appears downstream
of bottlenecks with congested traffic.\\
%
(iii) Non-stationary and non-homogeneous states
(see also Refs. \cite{Ker98b,CasBe99,TreHeHe00}).
For these, we will use the term {\em ``oscillating congested traffic''} (OCT)
\cite{HelHeTr99}.\\
At least types (i) and (iii) are characterized by a considerably scattering and erratic
change of the flow-density data, the various sources of which will be addressed
in the following subsection.
Continuous transitions between these types are probably the reason for the
so-called {\em ``pinch effect''} %explaining the formation of wide moving jams
\cite{Ker98a}, see Fig.~\ref{Fig3}(a):\par\begin{figure}[htbp]
\vspace*{-3mm}
\begin{center}\hspace*{-2mm}
\includegraphics[width=7cm]{fig3a.eps}\hspace*{-6mm}
\includegraphics[width=6cm]{fig3b.eps}
\end{center}
\caption[]{(a) Simulation of the pinch effect with a deterministic microscopic model
showing stable traffic at low and
high densities, linearly unstable traffic at medium densities, and metastable traffic
in between. The spatio-temporal density plot illustrates
the breakdown to homogeneous congested traffic (HCT) upstream of a bottleneck,
emerging oscillating congested traffic (OCT) further upstream, and a few stop-and-go waves
(TSG) emanating from this region. The conditions for this
spatial coexistence of congested traffic states are as follows \cite{Review,TreHe99b}:
The density in the congested region immediately
upstream of the bottleneck should be in the linearly unstable, but convectively stable
range, where perturbations are convected away in
upstream direction \cite{Mann90,CrosHo93}.
In this case, traffic flow will appear stationary and homogeneous close to the bottleneck,
but small perturbations will grow as they propagate upstream in the congested
region starting at the bottleneck. If the perturbations propagate faster than
the congested region expands, they will reach the area of free traffic upstream of the
bottleneck. During rush hours, it is quite likely that this free flow is in the metastable
range between the free and linearly unstable density region.
Consequently, sufficiently large perturbations
can trigger the formation of jams, which continue travelling upstream, while
small perturbations are absorbed. (b) The wavelength (average distance between
density maxima) determined from (a) qualitatively displays the empirical increase with the
vehicle velocity observed by Kerner \cite{Ker98a}.\label{Fig3}}
\end{figure}
{\noindent Upstream} of a section with
homogeneous congested traffic close to a bottleneck, there is a
so-called {\em ``pinch region''} characterized by the spontaneous
birth of small and narrow
density clusters, which are growing while they travel further
upstream. Wide moving jams are eventually formed by the merging or disappearance of
narrow jams, in which the cars move slower than in
wide jams \cite{Ker98a}. Once formed, the wide jams seem to suppress the occurence of new
narrow jams in between. Similar findings were already reported by Koshi {\em et al.} \cite{KosIwOh83},
who observed that ``ripples of speed grow larger
in terms of both height and length of the waves
as they propagate upstream''.
\subsection{Wide scattering of congested flow-density data}\label{widescat}
The collection and evaluation of empirical freeway data is a subject with often
underestimated problems. To make reliable conclusions, in original investigations
one should specify
%
(i) the measurement site and conditions (including applied control
measures),
%
(ii) the sampling interval,
%
(iii) the aggregation method,
%
(iv) the statistical properties
(variances, frequency distributions, correlations, survival times of traffic states, etc.),
%
(v) data transformations,
%
(vi) smoothing procedures,
%
and the respective dependencies on them.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\par\begin{figure}[htbp]
\begin{center}
\unitlength1cm
\begin{picture}(12,8)
\put(-0.25,4){\includegraphics[width=8.3cm]{fig4a.eps}}
\put(7.9,4.3){\includegraphics[width=4.2cm]{fig4b.eps}}
\put(-0.5,0.1){\includegraphics[width=4.6\unitlength]{fig4c.eps}}
\put(3.5,-0.3){\includegraphics[width=4.52\unitlength]{fig4d.eps}}
\put(7.7,-0.25){\includegraphics[width=4.12\unitlength]{fig4e.eps}}
\end{picture}
\end{center}
\caption[]{(a) The empirical truck fraction varies considerably in the course of time.
(b) The time headways of long vehicles (``trucks'') are on average much higher than
those of short vehicles (``cars''). (c)-(e) Assuming a fundamental diagram for cars (solid line), a separate
one for trucks (dashed line), weighting them according to the measured truck fraction, and
using empirical boundary conditions allows to reproduce the observations in a
(semi-)quantitatively way \cite{TreHe99a}: Free traffic (at low densities) is characterized by a
(quasi-)one-dimensional curve. (c) Data of congested traffic {\em upstream} of a bottleneck
are widely scattered in a two-dimensional area. (d) {\em Immediately downstream} of the bottleneck,
one observes homogeneous-in-speed states reflecting recovering traffic. (e) Further
downstream the data points approach the curve describing free traffic. Dark symbols correspond
to empirical one-minute data, light ones to corresponding simulation results.\label{Fig4}}
\end{figure}
The measurement conditions include ramps and road sections with their respective
in- and outflows,
speed limits, gradients, and curves with the respectively related capacities,
furthermore weather conditions (like rain, ice, blinding sun, etc.),
presence of incidents (including the opposite driving direction),
and other irregularities such as road works, which may
trigger a breakdown of traffic flow. Moreover, one
should evaluate the long vehicles (``trucks'') separately, as their proportion varies significantly,
see Fig.~\ref{Fig4}(a).
This can explain a considerable part of the wide scattering of congested traffic \cite{TreHe99a},
see Figs.~\ref{Fig4}(c), (d). Presently, this explanation is still the only one for this observation that
has been quantitatively checked with empirical data. Note that a considerable variation
of the time headways is also observed among cars, see Fig.~\ref{Fig4}(b).
This is partly due to different driver preferences
and partly due to the instability of traffic flow, see Fig.~\ref{Fig5}(a).
While vehicle platoons with reduced time headways imply an increase
of the flow with growing density, a reduction in vehicle speeds is usually related with a
decrease. According to Banks \cite{Bank99}, this can account for the observed erratic
changes of the flow-density data. We support this interpretation.
\par\begin{figure}[htbp]
\begin{center}
\begin{picture}(12,4.5)
\put(0,0.2){\includegraphics[width=6cm]{fig5a.eps}}
\put(5.9,0){\includegraphics[width=6.4cm]{fig5b.eps}}
\end{picture}
\end{center}
\caption[]{(a) The measured oscillations of the clearance %$s = (d - l)$
and the relative velocity \cite{Hoe72} %$\Delta v$ around $\Delta v = 0$ (cf. Hoefs, 1972)
indicate an instability in the car-following behavior \cite{HelTi98}.
%The existence of
%infinitely many stable congested flow-density states is, therefore, very questionable.
(b) The empirical standard deviation $\sqrt{\theta(t)}$ of vehicle velocities
divided by the average velocity $V(t)$ is particularly large during the rush hour, where
traffic flow is congested and unstable \cite{Review,Tilch01}.\label{Fig5}}
\end{figure}
The strong variations of traffic flows imply that all
measurements of macroscopic quantities should be complemented by
error bars (see, e.g., Ref.~\cite{Hal1AlGu86}).
Due to the relatively small ``particle'' numbers
behind the determination of macroscopic quantities, the error bars are
actually quite large. Hence, many temporal variations are within one
error bar, when traffic flow is unstable, see Fig.~\ref{Fig5}(b).
It is, therefore, very questionable whether it is possible
to empirically prove the existence of small perturbations triggering
a breakdown of traffic flow \cite{KerRe97} or of the ``birth'' and merging of narrow
density clusters in the ``pinch region'' \cite{Ker98a}. At least, this would require a
thorough statistical support. Consequently, we deny
that such kind of data are presently suited as starting point for the development of new models
\cite{KerMic} or traffic theories \cite{Kerner3}.
There is a considerable risk of overinterpreting particular (possibly statistical)
features of the data recorded at special freeway sections and to construct new models
that merely simulate what has been incorporated by means of the model assumptions.
In fact, the only reason why we believe in the correctness of these observations is the existence of
plausible deterministic traffic models reproducing these hard-to-see effects without
any special assumptions or extensions (see Fig.~\ref{Fig3} and
Refs.~\cite{Lee3LeKi98,TreHe99b,HelTr98a}).
\par\begin{figure}[htbp]
\begin{center}
\includegraphics[width=7cm]{fig6.eps}
\end{center}
\caption[]{\label{Fig6}The covariance between headways $d_\alpha$ and inverse
velocities $1/v_\alpha$
shows significant deviations from zero in congested traffic, while it approximately
vanishes in free flow, compare Fig.~\ref{Fig5}(b).
Even after traffic has recovered, there seem to remain
weak correlations between headways and vehicle speeds for a considerable time.
These are probably a reminiscence of congestion due to platoons which have
not fully dissolved \cite{Review,Tilch01}.}
\end{figure}
Because of the above mentioned problems, we would like to call for more refined measurement
techniques, which are required for more reliable data.
These must take into account {\em correlations} between different quantities, as is pointed out
by Banks \cite{Bank95}.
\par
For example, approximating the vehicle headways by $d_\alpha = v_\alpha \Delta t_\alpha$
(where $v_\alpha$ is the velocity and $\Delta t_\alpha$ the
time headway of vehicle $\alpha$)
and determining arithmetic multi-vehicle
averages $\langle \dots \rangle$ at a fixed location, one obtains for the inverse vehicle flow
\begin{equation}
\frac{1}{Q} = \langle \Delta t_\alpha \rangle
= \left\langle \! \frac{d_\alpha}{v_\alpha} \! \right\rangle
= \langle d_\alpha \rangle \left\langle \! \frac{1}{v_\alpha}
\! \right\rangle + \mbox{cov}\left(\!d_\alpha,\frac{1}{v_\alpha} \!\right) \, .
\end{equation}
Herein, $\mbox{cov}(d_\alpha,1/v_\alpha)$
denotes the covariance between the headways $d_\alpha$
and the inverse velocities $1/v_\alpha$, which
is negative and particularly relevant at large vehicle densities, as expected
(see Fig.~\ref{Fig6}). Defining the local density $\rho$ by
\begin{equation}
\rho = 1/\left\langle d_\alpha \right\rangle
\label{proposed}
\end{equation}
and the average velocity $V$ via
\begin{equation}
\frac{1}{V} = \left\langle \frac{1}{v_\alpha} \right\rangle \, ,
\label{harmav}
\end{equation}
we obtain the fluid-dynamic flow relation
\begin{equation}
Q = \rho V
\end{equation}
by the conventional assumption $\mbox{cov}(d_\alpha,1/v_\alpha) = 0$. This, however,
overestimates the density systematically, since the covariance tends to be negative
due to the speed-dependent safety distance of vehicles.
In contrast, the common method of determining the density via $Q/\langle v_\alpha\rangle$
systematically underestimates the density \cite{Review,TilHe00}.
{\em Consequently, errors in the measurement of the flow and the density due to
a neglection of correlations partly account for the observed scattering
of flow-density data in the congested regime.}
\par
A similar problem occurs when the density is determined via the time occupancy of
a certain cross section of the road. Considering that $\Delta t_\alpha
= T_\alpha + l_\alpha/v_\alpha$,
where $T_\alpha$ is the (netto) time clearance and $l_\alpha$
the length of vehicle $\alpha$, we have the relation
\begin{equation}
\rho = \rho_{\rm max} \frac{\langle l_\alpha/v_\alpha \rangle}{\langle \Delta t_\alpha \rangle}
= \rho_{\rm max} Q \langle l_\alpha / v_\alpha \rangle = \frac{Q}{\langle l_\alpha \rangle}
\langle l_\alpha / v_\alpha \rangle \, ,
\end{equation}
where $\rho_{\rm max} = 1/\langle l_\alpha \rangle$ is the maximum density and
$\langle l_\alpha \rangle$ the average vehicle length. For a finite detector length
$L_D$, we have to replace $l_\alpha$ by $l_\alpha+L_D$ \cite{Review,May90}. The formula $1/V =
\langle l_\alpha / v_\alpha \rangle/\langle l_\alpha \rangle$ for the average velocity results
in the correct expression $1/V = \langle 1/v_\alpha \rangle$ only,
if the individual vehicle lengths and velocities are not correlated, which is usually not the case.
\subsection{A quantitative theory of congested traffic states}
When Kerner started to question all traffic models with a fundamental diagram,
physicists were used to simulate traffic in closed systems with periodic boundary conditions.
With the Kerner-Konh\"auser model, it was possible to produce
free traffic, emergent stop-and-go waves, and triggered wide moving jams \cite{Kuh84a,Kuh91a,KerKo94}. However,
attempts to simulate ``synchronized flow'' failed even when small ramp flows were
added to the system. They resulted in what we call a {\em pinned localized cluster}
(PLC) located at the on-ramp \cite{KerPLC} (see Figs.~\ref{Fig9} and \ref{Fig10}).
Because of the sensitivity of the model
and problems with the treatment of open systems, it was not possible to simulate open
systems with considerable ramp flows. Other independent studies for periodic systems
with localized bottlenecks produced either homogeneous vehicle queues (HCT)
or {\em oscillating congested traffic} (OCT)
\cite{Lee3LeKi98,MunHsLa71,Phi77,Cre79,MakNaToMi81,ChuHu94,EmmRa95,CsaVi94,Hil95,KlaKuWe96,Naga97c},
but at that time nobody could make sense of these apparently inconsistent findings.
This situation changed, when Helbing {\em et al.} derived a phase diagram of
congested traffic states. They managed to simulate a macroscopic traffic model with open
boundary conditions even in extreme situations
and investigated a freeway stretch with a single ramp \cite{HelHeTr99}.
Instead of the densities, they identified the main flow on
the freeway and the on-ramp flow as the suitable control parameters for on open system
and varied them systematically. In this way, they found that a perturbation could trigger
different kinds of congested traffic states. Moreover, the boundaries separating different states
could be related to the {\em instability diagram} for homogeneous
freeways and other characteristic quantities \cite{HelHeTr99}. For this reason, they concluded that the
phase diagram should be qualitatively the same, i.e. {\em universal}, for all microscopic,
mesoscopic, or macroscopic traffic models having the same instability diagram. This
has been supported in the meantime \cite{HelHeTr99,TreHeHe00,Lee3LeKi99}.
Apart from this, the phase diagram of traffic models with
different instability diagrams can be directly derived \cite{Review}. Generalizations to other kinds
of bottlenecks (e.g. gradients) have been developed as well \cite{TreHeHe00}.
\par\begin{figure}[htbp]
\begin{center}
\includegraphics[height=.9\textwidth,angle=-90]{fig7.eps}\\[3mm]
\end{center}
\caption[]{Negative perturbation triggering oscillating congested traffic.
When the traffic density has sufficiently
increased to reach the metastable regime, the ``negative'' perturbation will
be amplified if it only exceeds the critical amplitude. While it is small, it will move
downstream with the vehicles, so one could hope it would pass the bottleneck and leave the system.
However, when the density wave grows larger, it will reduce its speed and even change its
propagation direction.
Once it is fully developed, it moves upstream with constant velocity, since vehicles
leave the jam at the downstream front, while new ones join it at the upstream front.
Hence, the perturbation returns to the bottleneck (see Fig.~\ref{Fig8} for this
{\em ``boomerang effect''}), and it triggers
a breakdown of traffic, when it passes the bottleneck in upstream direction, as
it thereby reduces the effective capacity to $Q_{\rm out}$
\cite{Review,HelTr98a,PhysikBl}.\label{Fig7}}
\end{figure}
In the following, we will sketch the basic ideas behind the phase diagram of congested
traffic states (for a more detailed discussion see Ref.~\cite{Review}). Let us assume our
traffic model has a fundamental diagram
\begin{equation}
Q_{\rm f}(\varrho) = \varrho V_{\rm f}(\varrho)
\end{equation}
describing the relation between the vehicle flow $Q$,
the {\em spatially averaged} density $\varrho$, and the average velocity $V$
in homogeneous and stationary traffic. (The flow-density relation of emergent
stop-and-go waves is characterized by a linear relation, i.e. it looks differently \cite{KerKo94}.)
Moreover, let us assume the model has ranges of stable traffic flow at small and high densities,
a range of linearly unstable traffic flow at medium densities, and ranges of metastable traffic flow
in between. This kind of instability diagram is, for example, found for the macroscopic model
used by K\"uhne, Kerner and Konh\"auser, or Lee {\em et al.} \cite{Lee3LeKi98,Kuh84a,KerKo94}, for
the microscopic optimal velocity model \cite{BandHaNaNaShSu95}, for the non-local gas-kinetic-based
traffic model \cite{TreHeHe99}, or the microscopic intelligent driver model \cite{TreHeHe00} (among which
the first two models are rather sensitive to parameter variations, but the latter two are
quite robust).
\par
In contrast to circular freeways, emergent ``phantom traffic jams'' are not typical for open
homogeneous freeway stretches, as it is normally impossible to reach the linearly unstable
density regime by feeding in vehicles at the upstream boundary. This is in agreement
with empirical observations \cite{Kerner4}.
Most cases of traffic congestion on an $n$-lane freeway
are observed upstream of on-ramps or other bottlenecks. They can be triggered by perturbations
significantly below the theoretical capacity, as soon as the sum of the upstream
freeway flow $Q_{\rm up}$ and the on-ramp flow $\Delta Q = Q_{\rm rmp}/n$ per lane exceeds the
outflow $Q_{\rm out}$ from congested traffic: If a disturbance leads to
temporary congestion, the drivers must accelerate again and suffer some time delay,
which reduces the capacity to $Q_{\rm out}$. Therefore, the following vehicles will
queue up, and the temporary perturbation grows to form a persistent kind of congestion.
The initial perturbation can even be a temporary {\em reduction} of the traffic flow and/or vehicle
density, which can be caused by temporal variations of the traffic volume or even by
vehicles leaving the freeway at some off-ramp \cite{DagCaBe99,PhysikBl}, see Fig.~\ref{Fig8}.
\par
\begin{figure}[htbp]
\begin{center}
\includegraphics[width=.8\textwidth]{fig8.eps}\\[3mm]
\end{center}
\caption[]{The wide moving jam left of kilometer 43 starts with a
``boomerang effect'' and travels through the ``synchronized'' congested traffic flow
left of kilometer 41 (dark area).
(Reproduction with kind permission of Rudolf Sollacher, Siemens AG, Munich.)
\label{Fig8}}
\end{figure}
\begin{figure}[htbp]
\begin{center}
\includegraphics[width=12cm]{fig9.eps}
\end{center}
\caption[]{Simulated representatives (density-over-space-and-time plots)
of the different congested traffic states, which were triggered by a big
perturbation travelling upstream. Center right: Phase diagram of the traffic states as a function
of the (upstream) traffic volume $Q_{\rm up}$ on the freeway and
the strength $\Delta Q$ of a bottleneck at location $x=0$~km, e.g. an on-ramp
with inflow $\Delta Q$ per freeway lane.
\label{Fig9}}
\end{figure}
If the total traffic volume $Q_{\rm tot} = (Q_{\rm up} + \Delta Q)$ is greater than the
dynamic capacity $Q_{\rm out}$, we will automatically end up with a growing vehicle queue
upstream of the on-ramp. The traffic flow $Q_{\rm cong}$ resulting in the congested
area normally gives, together with the inflow $\Delta Q$, the outflow $Q_{\rm out}$, i.e.
$Q_{\rm cong} = (Q_{\rm out} - \Delta Q)$ (but see the footnote on p. 1111 of Ref.~\cite{Review}
for exceptions). One can distinguish the following cases
\cite{HelHeTr99,TreHeHe00,Review} (see Fig.~\ref{Fig9}):
If the density $\varrho_{\rm cong}$ associated with the
flow $Q_{\rm cong} = Q_{\rm f}(\varrho_{\rm cong})$ is stable, we find {\em homogeneous
congested traffic} (HCT) such as typical traffic jams during holiday seasons.
For a smaller on-ramp flow $\Delta Q$, the congested flow $Q_{\rm cong}$ is linearly unstable,
and we either find {\em oscillating congested traffic} (OCT) or {\em triggered stop-and-go traffic}
(TSG), which often emerges from a spatial sequence of homogeneous and oscillating
congested traffic (so-called {\em ``pinch effect''} \cite{Ker98a}).
In contrast to OCT, stop-and-go
traffic is characterized by a sequence of moving jams, between which traffic flows freely.
Each traffic jam triggers the next one by inducing a small perturbation at the ramp,
which propagates downstream as long as it is small, but turns back when it has grown
large enough ({\em ``boomerang effect''}, cf.
Figs.~\ref{Fig7} to \ref{Fig10}). This, however, requires the downstream traffic flow
to be linearly unstable. If it is metastable instead (when the traffic volume is further
reduced), a traffic jam will usually not trigger a growing perturbation. In that case, one
finds either a single {\em moving localized cluster} (MLC), or a {\em pinned localized cluster} (PLC)
at the location of the ramp. The latter requires the traffic flow in the upstream section to be stable,
so that no traffic jam can survive there. Finally, for sufficiently small traffic volumes,
we find {\em free traffic} (FT), as expected.
\par%\unitlength0.68cm
%\begin{figure}[htbp]
%\begin{picture}(16,7)
%\put(0,3.3){{PLC}}
%\put(0,1.0){{TSG}}
%\put(1,0){\includegraphics[height=5.5\unitlength, angle=0]{fig9a.eps}}
%\put(16.7,4.3){{HCT}}
%\put(9,0){\includegraphics[height=5.5\unitlength, angle=0]{fig9b.eps}}
%\end{picture}
%\caption{Empirical representatives of homogeneous congested traffic (HCT),
%and of triggered stop-and-go traffic
%(TSG) combined with a pinned localized cluster (PLC) as observed on the
%Germany freeway A5 near Frankfurt.\label{Fig9}}
%\end{figure}
The different congested traffic states found
in the microsimulations (as displayed in Fig.~\ref{Fig9}) could all be identified in real traffic data
(see Fig.~\ref{Fig10} for some examples).
Moreover, according to our first investigation results,
the traffic patterns observed on the German freeway A5 near Frankfurt
have a typical dependence on the respective weekday and are even quantitatively
consistent with the phase diagram (see Fig.~\ref{Fig10}).
Of course, the empirically measured patterns look less regular, as the simulation results
displayed in Fig.~\ref{Fig9} are for a deterministic model with identical vehicle parameters.
\par\unitlength0.68cm
\begin{figure}[htbp]
\begin{picture}(16,16)
\put(0,9.5){\includegraphics[width=6cm]{fig10a.eps}}
\put(9,8){\includegraphics[width=6cm]{fig10b.eps}}
\put(-0.3,0.5){\includegraphics[width=6cm]{fig10c.eps}}
\put(9,0.5){\includegraphics[width=6cm]{fig10d.eps}}
\put(4,15.5){\footnotesize TSG}
%\put(16,12.5){\footnotesize PLC}
%\put(8,12.5){\footnotesize PLC}
\put(12.5,14){\footnotesize OCT}
%\put(16.7,11){\footnotesize PLC}
\put(16.3,6.5){\footnotesize PLC}
\put(17.1,4.5){\footnotesize PLC}
\end{picture}
\caption[]{Empirical examples of triggered stop-and-go traffic (TSG), oscillating
congested traffic (OCT), and pinned localized clusters (PLC), and the location of
the empirical data points in the phase diagram for the German freeway
A5 near Frankfurt.\label{Fig10}}
\end{figure}
We will now reply to some criticism and misunderstandings:
(1) Although the phase diagram and the congested traffic states have
been derived for identical driver-vehicle units and one bottleneck only, many observations
already fit very well into this scheme. If several bottlenecks are present,
the situation becomes more complicated, but can be addressed by similar methods.
In such cases, we may find the spatial {\em coexistence} of states such as OCT and PLC,
{\em transitions} between different states, extended congested traffic states (HCT, OCT, or TSG)
with a {\em fixed upstream front,} and other phenomena. The phenomenon of multistability and
coexisting states is, by the way, already found for the case of
one single bottleneck (see Ref. \cite{Lee3LeKi99} and Fig.~8 in Ref. \cite{TreHeHe00}).
(2) Because of a certain {\em ``penetration depth''} \cite{TreHeHe00}, MLC states can propagate
through small areas of stable traffic.
(3) The variation and scattering of the
flow-density data is well reproduced, if different driver-vehicle types are distinguished
(and overtaking maneuvers are taken into account).
(4) The ``pinch effect'' does not correspond to a transition between different phases
(HCT, OCT, and TSG) in the course of time, but it corresponds to a certain area in
the phase diagram, where the congested flow is convectively stable but linearly unstable
(not shown). The empirically observed increase of the vehicle velocity in ``synchronized
flow'' with the average oscillation wavelength \cite{Ker98a} is qualitatively
well reproduced, see Fig.~\ref{Fig3}(b). (5) The flow downstream of congested traffic
can exceed the dynamical capacity $Q_{\rm out}$, if
vehicles can enter the freeway via the ramp downstream of the congestion
front \cite{Kerner3}. This
is practially relevant for particularly long on-ramps like freeway intersections
(see Footnote on p.~1111 of Ref.~\cite{Review}).
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\section{Pedestrian Evacuation}
\unitlength0.75cm
The topics of pedestrian traffic and evacuation of buildings, stadia, and ships have recently
attracted great interest among traffic scientists. Here, we will give a short overview only, as
there are several detailed reviews available (see Refs.~\cite{Hel97a,Review,HelMo2FaBo00,SchSh01,PED}).
We will focus on
the {\em social-force model} of pedestrian dynamics which describes the different
competing motivations of pedestrians by separate force terms. It has the following advantages:\\
(1) The social-force model takes into account the flexible usage of space
(i.e. the compressibility of the crowd), but also the
excluded volume and friction effects which play a role at extreme densities.
(2) The model assumptions are simple and plausible. Virtual fields \cite{BurKlaSchZi01}
or other questionable model ingredients are not necessary to obtain realistic results.
(3) There are only a few model parameters to calibrate.
(4) The model is robust and naturally reproduces many different observations without
modifications of the model.
(5) Nevertheless, it is easy to consider individual
differences in the behavior, and extensions for more complex problems such as
trail formation \cite{Hel97a,HelMo2FaBo00,HelKe1Mo97} are possible.
\par
\unitlength0.75cm
\begin{figure}
\begin{center}
\begin{picture}(12,6.8)
\put(-0.05,3.5){\includegraphics[width=12.1\unitlength]{fig11a.eps}}
\put(0,0){\includegraphics[width=12\unitlength]{fig11b.eps}}
\put(-1,6){{\footnotesize (a)}}
\put(-1,2.5){{\footnotesize (b)}}
\end{picture}
\end{center}
\caption{(a) Formation of lanes in initially disordered pedestrian crowds
with opposite walking directions and small noise amplitudes $\eta_i$
(after \cite{Review,PED,HelFaVi00a}; cf. also
\cite{HelMo295,HelVi99}).
White disks represent pedestrians moving
from left to right, black ones move the other way round.
(b) For sufficiently high densities and large fluctuations, we observe
the noise-induced formation of a crystallized, ``frozen'' state
(after \cite{Review,HelFaVi00a}).\label{Fig11}}
\end{figure}
For normal, ``relaxed'' situations with small fluctuation amplitudes,
our microsimulations of counterflows in corridors reproduce the
empirically observed {\em formation of lanes} consisting of pedestrians with the
same desired walking direction \cite{Review,HelFaVi00a,HelMo295,HelVi99}, see Fig.~\ref{Fig11}(a).
If we do not assume periodic boundary conditions, these lanes are dynamically varying
(see the Java applet at {\tt www.helbing. org/Pedestrians/ Corridor.html}). Their number
depends on the width of the street \cite{HelMo295},
on pedestrian density, and on the noise level. Interestingly, one finds
a {\em noise-induced ordering} \cite{HelPl00}:
Compared to small noise amplitudes, medium ones result in a more pronounced segregation
(i.e., a smaller number of lanes). Large noise amplitudes lead to
a {\em ``freezing by heating''} effect characterized by a breakdown of ``fluid''
lanes and the emergence of ``solid'' blockages \cite{HelFaVi00a}, see Fig.~\ref{Fig11}(b).
Note that our model can explain lane formation even without assuming asymmetrical interactions
or attraction effects \cite{HelFaVi00a,HelMo295,HelVi99}.
It is an {\em optimal self-organization} phenomenon \cite{HelVi99}
resulting from the combination of driving and repulsive
forces. The same model also reproduces the
observed oscillations of the flow direction at bottlenecks \cite{Review,PhysikBl,HelMo295}
without the need of a virtual ``floor field'' \cite{Scha01,BurKirKlaSchZi01}. {\em Cellular automaton}
Java applets from 1998 are available in the internet to visualize these phenomena
(see {\tt www.helbing.org/Pedestrians/ Corridor.html, Door.html}). They are based on
a discretization of the social force model that can be viewed as a discrete two-dimensional
optimal velocity model \cite{Bolay98}. For other cellular automata see
Ref.~\cite{SchSh01,BlueAd98,FukIs99a,MuraIrNa99,KluMeWaSc00}.
\par
``Freezing by heating'' is one of the phenomena observed in crowd stampedes. Another
one is the {\em ``faster-is-slower effect'' (or ``slower-is-faster effect'')}
\cite{HelFaVi00b}. It can trigger a {\em ``phantom panic''} \cite{HelFaVi00b} and is caused
by arching and clogging at bottlenecks like exits, see Fig.~\ref{Fig12}(a).
\begin{figure}[htbp]
\begin{picture}(17,12.1)
\put(0.4,6){\includegraphics[height=6\unitlength,clip=]{fig12a.eps}}
\put(10,11.9){\includegraphics[width=5.6\unitlength, angle=-90]{fig12b.eps}}
\put(-0.1,11.3){{\footnotesize (a)}}
\put(9.95,11.3){{\footnotesize (b)}}
\put(0.1,0.5){\includegraphics[width=7.4\unitlength,clip=]{fig12c.eps}}
\put(8.1,0.5){\includegraphics[width=7.4\unitlength,clip=]{fig12d.eps}}
\put(0.1,5.5){\footnotesize (c)}
\put(8.1,5.5){\footnotesize (d)}
\end{picture}
\caption[]{(a)When the desired velocity $v_0$ is too high (e.g. in panic situations),
pedestrians come so close to each other, that their
physical contacts cause the build up of pressure and obstructing friction effects,
which results in temporary arching and clogging (for online Java simulations see
{\tt http:/$\!$/angel.elte.hu/$\tilde{\hphantom{n}}$panic/}).
(b) This is related with an
inefficient and irregular outflow, while the outflow is regular
for small enough desired velocities ($v_0 \le 1.5$~m/s) \cite{Review,PED,HelFaVi00b}.
(c) In panicking crowds, high pressures build up due to physical interactions.
This can injure people (black disks), who turn into obstacles for other pedestrians trying to leave
\cite{PED}.
(d) A column in front of the exit (large black disk) can avoid injuries by taking up pressure
from behind. It can also significantly increase the outflow \cite{PED}. In large exit areas
used by several hundret people,
several randomly placed columns are needed to subdivide the crowd and the pressure. An asymmetric
configuration of the columns is most efficient, as it avoids equilibria of forces which may
temporarily stop the outflow. (See {\tt http:/$\!$/angel.elte.hu/$\tilde{\hphantom{n}}$panic/}
for online Java simulations.)\label{Fig12} }
\end{figure}
The underlying reason is
the friction effect occuring in dense crowds, if the desired velocity is so high that
pedestrians have physical interactions. In these situations, extreme pressure can build
up in the crowd, and people may be crushed or trampled, in this way turning into
obstacles for other fleeing people, see Fig.~\ref{Fig12}(c).
These dangerous pressures can be reduced by columns, when suitably placed
in front of the exits, see Fig.~\ref{Fig12}(d). Thereby, the number of injuries can be
significantly reduced, and the outflows are considerably increased (see
the Java applets at {\tt http:/$\!$/angel.elte.hu/$\tilde{\hphantom{n}}$panic/}).
\par
Not only bottlenecks are dangerous in panic situations, but also widenings \cite{HelFaVi00b}. These
can reduce the efficiency of leaving, see Figs.~\ref{Fig13}(a), (b). Another problem is
{\em herding behavior}, as it is responsible for an inefficient use of available exits \cite{Review,HelFaVi00b},
see Figs.~\ref{Fig13}(c), (d).
\par
\begin{figure}
\begin{center}
\begin{picture}(16,12)
\put(0,6){\includegraphics[height=5.8\unitlength,clip=]{fig13a.eps}}
\put(10,11.7){\includegraphics[width=5.6\unitlength, angle=-90]{fig13b.eps}}
\put(0,10.2){\footnotesize (a)}
\put(11.6,10.7){\footnotesize (b)}
\put(-0.0,5.4){\includegraphics[height=4.35\unitlength, angle=180]{fig13c.eps}}
\put(7.9,5.7){\includegraphics[width=5.45\unitlength, angle=-90]{fig13d.eps}}
\put(0,5.7){\footnotesize (c)}
\put(7.85,5.7){\footnotesize (d)}
\end{picture}
\end{center}
\caption[]{(a), (b) Drop of the efficiency of leaving in corridors with widenings \cite{PED,HelFaVi00b}.
(c), (d) Herding behavior of panicking pedestrians in a smoky room (black),
leading to an inefficient use of
available escape routes \cite{PED,HelFaVi00b}. (Online Java simulations of this effect are
provided at {\tt http:/$\!$/angel.elte.hu/$\tilde{\hphantom{n}}$panic/}.)\label{Fig13}}
\end{figure}
\section{Summary and Application to Optimal Production Processes}
Nowadays, most aspects of traffic dynamics have been understood on the basis of
self-driven many-particle models. The observed phenomena can be \mbox{(semi-)}quan\-ti\-ta\-tively
reproduced by simulations using measured boundary conditions \cite{TreHeHe00}. Moreover, a
universal phase diagram of congested traffic states for freeway sections with
one bottleneck has been found, and the generalization to more complex situations
is straightforward. The reproduction of fine details, however, will require a more detailed
measurement of the interactive vehicle dynamics and the consideration of psychological
aspects. Although these may also be described in a mathematical way \cite{KerMic}, it will be hardly
possible to prove or disprove the corresponding models, i.e. the criteria demanded in
the natural sciences would have to be relaxed. A more promising research direction
is the modelling and optimization of production processes.
%, as many conclusions from traffic modelling can be transfered.
For example, applying the knowledge of the ``slower-is-faster effect'' to the treatment times in
a series of production steps, we were able to increase the throughput of
production processes in a major Chip factory by up to 39\% \cite{Fasold}.
\par
\begin{figure}[htbp]
\unitlength0.8cm
\begin{center}
\begin{picture}(14,14)
\put(0.2,14.5){\includegraphics[width=7.3\unitlength,angle=-90]{fig14a.eps}}
\put(0,7){\includegraphics[width=7\unitlength,angle=-90]{fig14b.eps}}
\put(0,14.3){\footnotesize (a)}
\put(0,6.7){\footnotesize (b)}
\end{picture}
\end{center}
\caption[]{(a) Gantt diagramm illustrating the treatment times in
different modules, corresponding to different production steps. The limiting
factor for increasing the throughput is the transport module.
(b) Applying the knowledge of the ``slower-is-faster effect'' to the treatment times,
we were able to increase the throughput of
some production processes in a major Chip factory by 33\% or even
more \cite{Fasold}.\label{Fig14}}
\end{figure}
Many conclusions from traffic models are relevant for the organization of societies,
administrations, companies, production processes, and so on, as the basic model ingredients
agree:
(1) The system consists of a large number of similar {\em units/entities}
(individuals, pedestrians, cars, boxes, ...).
(2) The units are externally or internally {\em driven,} i.e. there is some energy input, e.g.,
they can move.
(3) Units are {\em non-linearly interacting,} i.e. under certain conditions,
small variations can have large effects.
The system behavior is dominated by interactions rather than by the
boundary conditions (the external control).
(4) There is a {\em competition for resources} such as time (slots),
space, money.
(5) Each unit has a certain {\em extension} in space or time.
(6) When units come too close, they have {\em frictional effects} and obstruct each other.
\begin{thebibliography}{99}
\addcontentsline{toc}{section}{References}
\bib{KerRe96b}
B. S. Kerner, H. Rehborn: %, 1996b,
\tit{Experimental properties of complexity in traffic flow}
\art{Phys. Rev. E}{53}{R4275}{R4278}{(1996)}
\bib{FOTO}
B. S. Kerner, H. Rehborn, M. Aleksic, A.Haug, R. Lange: %, 2000c,
\tit{Verfolgung und Vorhersage von Verkehrsst\"orungen auf Autobahnen mit
`ASDA' und `FOTO' im online-Betrieb in der Verkehrsrechnerzentrale R\"usselsheim}
\art{Stra{\ss}enverkehrstechn.}{10}{521}{527}{(2000)}
\bib{KerReAl00}
B. S. Kerner, H. Rehborn, M. Aleksic: %, 2000b,
\tit{Forecasting of traffic congestion}
in {\em Traffic and Granular Flow '99}, edited by
D. Helbing, H. J. Herrmann, M. Schreckenberg, and D. E. Wolf (Springer, Berlin 2000), pp. 339--344
\bib{NeubSaScSc99}
L. Neubert, L. Santen, A. Schadschneider, M. Schreckenberg: %, %1999{b},
\tit{Single-vehicle data of highway traffic: A statistical analysis}
\art{Phys. Rev. E}{60}{6480}{6490}{(1999)}
\bib{MikaKrYu69}
H. S. Mika, J. B. Kreer, L. S. Yuan: %, 1969,
\tit{Dual mode behavior of freeway traffic}
\art{Highw. Res. Rec.}{279}{1}{13}{(1969)}
\bib{KerRe96a}
B. S. Kerner, H. Rehborn: %, 1996a,
\tit{Experimental features and characteristics of traffic jams}
\art{Phys. Rev. E}{53}{R1297}{R1300}{(1996)}
\bib{CasMa99}
M. J. Cassidy, M. Mauch: %, 2001,
\tit{An observed traffic pattern in long freeway queues}
\art{Transpn. Res. A}{35}{143}{156}{(2001)}
\bib{Kerner1}
B. S. Kerner, S. L. Klenov, P. Konh\"auser: %, 1997,
\tit{Asymptotic theory of traffic jams}
\art{Phys. Rev. E}{56}{4200}{4216}{(1997)}
\bib{Per86}
B. N. Persaud: %, 1986,
\disstit{Study of a freeway bottleneck to explore some unresolved traffic
flow issues}
Ph.D. thesis, University of Toronto (1986) % Canada
\bib{Bank90}
J. H. Banks: %, 1990,
\tit{Flow processes at a freeway bottleneck}
\art{Transpn. Res. Rec.}{1287}{20}{28}{(1990)}
\bib{Bank91a}
J. H. Banks: %, J. H., 1991a,
\tit{Two-capacity phenomenon at freeway bottlenecks: A basis for ramp
metering?}
\art{Transpn. Res. Rec.}{1320}{83}{90}{(1991)}
%
%\bib{AgyHa91}
%Agyemang-Duah, K., and F.~L. Hall, 1991,
%\tit{Some issues regarding the numerical value of freeway capacity}
%in {\em Highway Capacity and Level of Service}, edited by U. Brannolte
%(Balkema, Rotterdam), \pag{1}{15}
\bib{KerKirReh}
B. S. Kerner, H. Kirschfink, H. Rehborn: %, 1999,
\tit{Method for the automatic monitoring of traffic including the analysis of back-up dynamics}
German Patent DE 196 47 127.3; US Patent US 5,861,820 (1999)
\bib{LighWh55}
M. J. Lighthill, G. B. Whitham: %, 1955,
\tit{On kinematic waves: {II. A} theory of traffic on long crowded roads}
\art{Proc. Roy. Soc. London, Ser. A}{229}{317}{345}{(1955)}
\bib{KosIwOh83}
M. Koshi, M. Iwasaki, I. Ohkura: %, 1983,
\tit{Some findings and an overview on vehicular flow characteristics}
In: {\it Proceedings of the 8th International Symposium on
Transportation and Traffic Flow Theory},
ed. by V. F. Hurdle, E. Hauer, G.~N. Stewart (University of Toronto, Toronto, Ontario 1983) \pag{403}{426}
\bib{EdiFo58}
L. C. Edie, R. S. Foote: %, 1958,
\tit{Traffic flow in tunnels}
\art{High. Res. Board Proc.}{37}{334}{344}{(1958)}
\bib{ForMuSi67}
T. W. Forbes, J. J. Mullin, M. E. Simpson: %, 1967,
\tit{Interchange spacings and driver behavior effects on freeway
operations}
In: {\em Proceedings of the 3rd International Symposium on the Theory
of Traffic Flow}, ed. by L. C. Edie (Elsevier, New York, N. Y. 1967)
\pag{97}{108}
\bib{Lee3LeKi98}
H. Y. Lee, H.-W. Lee, D. Kim: %, 1998,
\tit{Origin of synchronized traffic flow on highways and its dynamic
phase transitions}
\art{Phys. Rev. Lett.}{81}{1130}{1133}{(1998)}
\bib{KerRe97}
B. S. Kerner, H. Rehborn: %, 1997,
\tit{Experimental properties of phase transitions in traffic flow}
\art{Phys. Rev. Lett.}{79}{4030}{4033}{(1997)}
\bib{ShvHe99}
V. Shvetsov, D. Helbing: %, 1999,
\tit{Macroscopic dynamics of multilane traffic}
\art{Phys. Rev. E}{59}{6328}{6339}{(1999)}
\bib{HelHu98}
D. Helbing, B. A. Huberman: %, 1998,
\tit{Coherent moving states in highway traffic}
\art{Nature}{396}{738}{740}{(1998)}
\bib{Kerner3}
B. S. Kerner: % 2001,
\tit{Complexity of synchronized flow and related problems for basic assumptions
of traffic flow theories.}
\art{Networks and Spatial Economics}{1}{35}{76}{(2001)}
\bib{Hel97a}
D. Helbing: % 1997a,
{\em Verkehrsdynamik}
(Springer, Berlin 1997)
\bib{Hel97c}
D. Helbing: % 1997c,
\tit{Fundamentals of traffic flow}
\art{Phys. Rev. E}{55}{3735}{3738}{(1997)}
%
%\bib{TreitTa66}
%Treiterer, J., and J.~I. Taylor, 1966,
%{\em Traffic Flow Investigations by Photogrammetric Techniques}
%\art{Highw. Res. Rec.}{142}{1}{12}
\bib{TreitMy74}
J. Treiterer, J. A. Myers: %, 1974,
\tit{The hysteresis phenomenon in traffic flow}
In: {\em Proceedings of the 6th International Symposium on Transportation
and Traffic Theory}, ed. by D. Buckley (Reed, London 1974) \pag{13}{38}
\bib{Pay84}
H. Payne: %, 1984,
\tit{Discontinuity in equilibrium traffic flow}
\art{Transpn. Res. Rec.}{971}{140}{146}{(1984)}
\bib{Hal187}
F.~L. Hall: % 1987,
\tit{An interpretation of speed-flow-concentration relationships using catastrophy theory}
\art{Transpn. Res. A}{21}{191}{201}{(1987)}
\bib{Hal1Ag91}
F. L. Hall, K. Agyemang-Duah: %, 1991,
\tit{Freeway capacity drop and the definition of capacity}
\art{Transpn. Res. Rec.}{1320}{91}{108}{(1991)}
\bib{PerYaBr98}
B. Persaud, S. Yagar, R. Brownlee: %, 1998,
\tit{Exploration of the breakdown phenomenon in freeway traffic}
\art{Transpn. Res. Rec.}{1634}{64}{69}{(1998)}
\bib{Wes98}
D. Westland: %, 1998,
\tit{Effect of highway geometry on freeway queuing at merge sections}
In: {\em Proceedings of the 3rd International Symposium on Highway Capacity},
ed. by R. Rysgaard (Road Directorate, Denmark 1998) \pag{1095}{1116}
\bib{HelHeTr99}
D. Helbing, A. Hennecke, M. Treiber: %, 1999,
\tit{Phase diagram of traffic states in the presence of inhomogeneities}
\art{Phys. Rev. Lett.}{82}{4360}{4363}{(1999)}
\bib{Ker98b}
B. S. Kerner: % 1998b,
\tit{A theory of congested traffic flow}
In: {\em Proceedings of the 3rd International Symposium on Highway Capacity},
ed. by R. Rysgaard (Road Directorate, Denmark 1998) Vol. 2, \pag{621}{642}
\bib{Lee3LeKi00}
H. Y. Lee, H.-W. Lee, D. Kim: %, 2000,
\tit{Phase diagram of congested traffic flow: An empirical study}
\art{Phys. Rev. E}{62}{4737}{4741}{(2000)}
\bib{Review}
D. Helbing: %, 2001,
\tit{Traffic and related self-driven many-particle systems}
\art{Reviews of Modern Physics}{1067}{1141}{(2001)}
\bib{CasBe99}
M. J. Cassidy, R. L. Bertini: %, 1999,
\tit{Some traffic features at freeway bottlenecks}
\art{Transpn. Res. B}{33}{25}{42}{(1999)}
\bib{TreHeHe00}
M. Treiber, A. Hennecke, D. Helbing: %, 2000,
\tit{Congested traffic states in empirical observations and microscopic simulations}
\art{Phys. Rev. E}{62}{1805}{1824}{(2000)}
\bib{Ker98a}
B. S. Kerner: % 1998a,
\tit{Experimental features of self-organization in traffic flow}
\art{Phys. Rev. Lett.}{81}{3797}{3800}{(1998)}
\bib{TreHe99b}
M. Treiber, D. Helbing: %, 1999b,
\tit{Explanation of observed features of self-organization in traffic flow}
e-print cond-mat/9901239 (1999)
\bib{Mann90}
P. Manneville: % 1990,
{\it Dissipative Structures and Weak Turbulence}
(Academic, New York 1990).
\bib{CrosHo93}
M.~C. Cross, P. C. Hohenberg: %, 1993,
\tit{Pattern formation outside of equilibrium}
\art{Rev. Mod. Phys.}{65}{851}{1112}{(1993)}
\bib{TreHe99a}
M. Treiber, D. Helbing: %, 1999a,
\tit{Macroscopic simulation of widely scattered synchronized traffic states}
\art{J. Phys. A: Math. Gen.}{32}{L17}{L23}{(1999)}
\bib{Bank99}
J. H. Banks: % 1999,
\tit{An investigation of some characteristics of congested flow}
\art{Transpn. Res. Rec.}{1678}{128}{134}{(1999)}
\bib{Hal1AlGu86}
F.~L. Hall, B.~L. Allen, M.~A. Gunter: %, 1986,
\tit{Empirical analysis of freeway flow-density relationships}
\art{Transpn. Res. A}{20}{197}{210}{(1986)}
\bib{KerMic}
B. S. Kerner, S. L. Klenov,
\tit{A micropscopic model for phase transitions in traffic flow}
\art{J. Phys. A: Math. Gen.}{35}{L31}{L43}{(2002)}
\bib{HelTr98a}
D. Helbing, M. Treiber: %, 1998a,
\tit{Gas-kinetic-based traffic model explaining observed hysteretic
phase transition}
\art{Phys. Rev. Lett.}{81}{3042}{3045}{(1998)}
\bib{Bank95}
J. H. Banks: %, 1995,
\tit{Another look at a priori relationships among traffic flow
characteristics}
\art{Transpn. Res. Rec.}{1510}{1}{10}{(1995)}
\bib{Hoe72}
D.~H. Hoefs: % 1972,
{\it Untersuchung des {F}ahrverhaltens in Fahrzeugkolonnen}
(Bundesministerium f{\"u}r Verkehr, Abt. Stra{\ss}enbau, Bonn-Bad Godesberg 1972)
%, Heft~140 der Reihe {\it Stra{\ss}enbau und Stra{\ss}enverkehrstechnik}.
\bib{HelTi98}
D. Helbing, B. Tilch: %, 1998,
\tit{Generalized force model of traffic dynamics}
\art{Phys. Rev. E}{58}{133}{138}{(1998)}
\bib{Tilch01}
B. Tilch: %, 2001,
\disstit{Modellierung und Simulation selbstgetriebener Vielteilchensysteme mit Anwendung
auf den Stra{\ss}enverkehr}
Ph.D. thesis, University of Stuttgart, in preparation
\bib{TilHe00}
B. Tilch, D. Helbing: %, 2000,
\tit{Evaluation of single vehicle data in dependence of the vehicle-type, lane, and site}
In: {\em Traffic and Granular Flow '99}, ed. by
D. Helbing, H. J. Herrmann, M. Schreckenberg, D. E. Wolf (Springer, Berlin 2000) \pag{333}{338}
\bib{May90}
A.~D. May: %, 1990,
{\em Traffic Flow Fundamentals}
(Prentice Hall, Englewood Cliffs, NJ 1990)
\bib{Kuh84a}
R.~D. K{\"u}hne: %, 1984a,
\tit{Macroscopic freeway model for dense traffic---{S}top-start
waves and incident detection}
In: {\em Proceedings of the 9th International Symposium on Transportation
and Traffic Theory}, ed. by
I. Volmuller, R. Hamerslag (VNU Science, Utrecht 1984) \pag{21}{42}
\bib{Kuh91a}
R.~D. K{\"u}hne: %, 1991a,
\tit{Traffic patterns in unstable traffic flow on freeways}
In: {\em Highway Capacity and Level of Service}, Proceedings of the International Symposium on Highway Capacity,
%{K}arlsruhe, 24-27 July 1991,
ed. by U. Brannolte (Balkema, Rotterdam 1991) \pag{211}{223}
%\bib{KerKo93}
%Kerner, B.~S., and P. Konh{\"a}user, 1993,
%\tit{Cluster effect in initially homogeneous traffic flow}
%\art{Phys. Rev. E}{48}{R2335}{R2338}
\bib{KerKo94}
B. S. Kerner, P. Konh\"auser: %, 1994,
\tit{Structure and parameters of clusters in traffic flow}
\art{Phys. Rev. E}{50}{54}{83}{(1994)}
\bib{KerPLC}
B. S. Kerner, P. Konh\"auser, M. Schilke: %, 1995a,
\tit{Deterministic spontaneous appearance of traffic jams in slightly
inhomogeneous traffic flow}
\art{Phys. Rev. E}{51}{R6243}{R6246}{(1995)}
\bib{MunHsLa71}
P.~K. Munjal, Y.-S. Hsu, R.~L. Lawrence: %, 1971,
\tit{Analysis and validation of lane-drop effects on multi-lane freeways}
\art{Transpn. Res.}{5}{257}{266}{(1971)}
\bib{Phi77}
W.~F. Phillips: %, 1977,
{\it Kinetic {M}odel for {T}raffic {F}low}
(National Technical Information Service, Springfield, VA 22161),
Technical Report~DOT/RSPD/DPB/50-77/17 (1977)
\bib{Cre79}
M. Cremer: %, 1979,
{\em Der Verkehrsflu{\ss} auf Schnellstra{\ss}en}
(Springer, Berlin 1979)
\bib{MakNaToMi81}
Y. Makigami, T. Nakanishi, M. Toyama, R. Mizote: %, 1983,
\tit{On a simulation model for the traffic stream in a freeway merging area}
In: {\it Proceedings of the 8th International Symposium on
Transportation and Traffic Flow Theory}, ed. by V. F. Hurdle, E. Hauer, G.~N. Stewart
(University of Toronto, Toronto, Ontario 1983) \pag{427}{451}
\bib{ChuHu94}
K. H. Chung, P. M. Hui: %, 1994,
\tit{Traffic flow problems in one-dimensional inhomogeneous media}
\art{J. Phys. Soc. Jpn.}{63}{4338}{4341}{(1994)}
\bib{EmmRa95}
H. Emmerich, E. Rank: %, 1995,
\tit{Investigating traffic flow in the presence of hindrances by cellular automata}
\art{Physica A}{216}{435}{444}{(1995)}
\bib{CsaVi94}
Z. Csah\'ok, Z., T. Vicsek: %, 1994,
\tit{Traffic models with disorder}
\art{J. Phys. A: Math. Gen.}{27}{L591}{L596}{(1994)}
\bib{Hil95}
M. Hilliges: % 1995,
{\it Ein ph{\"a}nomenologisches Modell des dynamischen Verkehrsflusses in Schnellstra{\ss}ennetzen}
(Shaker, Aachen 1995)
\bib{KlaKuWe96}
A. Klar, R.~D. K{\"u}hne, R. Wegener: %, 1996,
\tit{Mathematical models for vehicular traffic}
\art{Surv. Math. Ind.}{6}{215}{239}{(1996)}
\bib{Naga97c}
T. Nagatani: %, 1997c,
\tit{Instability of a traffic jam induced by slowing down}
\art{J. Phys. Soc. Jpn.}{66}{L1928}{L1931}{(1997)}
\bib{Lee3LeKi99}
H. Y. Lee, H.-W. Lee, D. Kim: %, 1999,
\tit{Dynamic states of a continuum traffic equation with on-ramp}
\art{Phys. Rev. E}{59}{5101}{5111}{(1999)}
\bib{BandHaNaNaShSu95}
M. Bando, K. Hasebe, K. Nakanishi, A. Nakayama, A. Shibata, Y. Sugiyama: %, 1995b,
\tit{Phenomenological study of dynamical model of traffic flow}
\art{J. Phys. I France}{5}{1389}{1399}{(1995)}
\bib{TreHeHe99}
M. Treiber, A. Hennecke, D. Helbing: %, 1999,
\tit{Derivation, properties, and simulation of a gas-kinetic-based,
non-local traffic model}
\art{Phys. Rev. E}{59}{239}{253}{(1999)}
\bib{Kerner4}
B. S. Kerner: %, 2000,
\tit{Experimental features of the emergence of moving jams in free traffic flow}
\art{J. Phys. A.: Meth. Gen.}{33}{L221}{L228}{(2000)}
\bib{DagCaBe99}
C. F. Daganzo, M. J. Cassidy, R. L. Bertini: %, 1999,
\tit{Possible explanations of phase transitions in highway traffic}
\art{Transpn. Res. A}{33}{365}{379}{(1999)}
\bib{PhysikBl}
D. Helbing: %, 2001,
\tit{Die wundervolle Welt aktiver Vielteilchensysteme}
\art{Phys. Bl.}{57}{27}{33}{(2001)}
\bib{HelMo2FaBo00}
D. Helbing, P. Moln\'{a}r, I. Farkas, K. Bolay: %, 2001c,
\tit{Self-organizing pedestrian movement}
\art{Env. Plan. B}{28}{361}{383}{(2001)}
\bib{SchSh01}
M. Schreckenberg, S. D. Sharma, Eds.: %, 2001,
{\em Pedestrian and Evacuation Dynamics}
(Springer, Berlin 2002)
\bibitem{PED}
D. Helbing, I. J. Farkas, P. Moln\'{a}r, T. Vicsek: %, 2001,
\tit{Simulation of pedestrian crowds in normal and evacuation situations}
In: {\em Pedestrian and Evacuation Dynamics}, ed. by M. Schreckenberg, S. D. Sharma
(Springer, Berlin 2002) \pag{21}{58}
\bib{BurKlaSchZi01}
C. Burstedde, K. Klauck, A. Schadschneider, J. Zittartz: %, 2001b,
\tit{Simulation of pedestrian dynamics using a 2-dimensional cellular automaton}
\art{Physica A}{295}{507}{525}{(2001)}
\bib{HelKe1Mo97}
D. Helbing, J. Keltsch, P. Moln\'{a}r: %, 1997{a},
\tit{Modelling the evolution of human trail systems}
\art{Nature}{388}{47}{50}{(1997)}
\bib{HelFaVi00a}
D. Helbing, I. Farkas, T. Vicsek: %, 2000a,
\tit{Freezing by heating in a driven mesoscopic system}
\art{Phys. Rev. Lett.}{84}{1240}{1243}{(2000)}
\bib{HelMo295}
D. Helbing, P. Moln\'{a}r: %, 1995,
\tit{Social force model of pedestrian dynamics}
\art{Phys. Rev. E}{51}{4282}{4286}{(1995)}
\bib{HelVi99}
D. Helbing, T. Vicsek: %, 1999,
\tit{Optimal self-organization}
\art{New J. Phys.}{1}{13.1}{13.17}{(1999)}
\bib{HelPl00}
D. Helbing, T. Platkowski: %, 2000,
\tit{Self-organization in space and induced by fluctuations}
\art{Int. J. Chaos Theor. Appl.}{5}{47}{62}{(2000)}
\bib{Scha01}
A. Schadschneider: %, 2001,
\tit{Cellular automaton approach to pedestrian dynamics---Theory}
In: {\em Pedestrian and Evacuation Dynamics}, ed. by M. Schreckenberg, S. D. Sharma
(Springer, Berlin 2002) \pag{75}{86}
\bib{BurKirKlaSchZi01}
C. Burstedde, A. Kirchner, K. Klauck, A. Schadschneider, J. Zittartz: %, 2001a,
\tit{Cellular automaton approach to pedestrian dynamics---Applications}
In: {\em Pedestrian and Evacuation Dynamics}, ed. by M. Schreckenberg, S. D. Sharma
(Springer, Berlin 2002) \pag{87}{98}
\bib{Bolay98}
K. Bolay: %, 1998,
\tit{Nichtlineare Ph\"anomene in einem fluid-dynamischen Verkehrsmodell}
%[``Non-Linear Phenomena in a Fluid-Dynamic Traffic Model'']}
Master's thesis, University of Stuttgart (1998)
\bib{BlueAd98}
V. J. Blue, J. L. Adler: %, 1998,
\tit{Emergent fundamental pedestrian flows from cellular automata microsimulation}
\art{Trans. Res. Rec.}{1644}{29}{36}{(1998)}
\bib{FukIs99a}
M. Fukui, Y. Ishibashi: %, 1999a,
\tit{Self-organized phase transitions in cellular automaton models for pedestrians}
\art{J. Phys. Soc. Jpn.}{68}{2861}{2863}{(1999)}
\bib{MuraIrNa99}
M. Muramatsu, T. Irie, T. Nagatani: %, 1999,
\tit{Jamming transition in pedestrian counter flow}
\art{Physica A}{267}{487}{498}{(1999)}
\bib{KluMeWaSc00}
H. Kl\"upfel, M. Meyer-K\"onig, J. Wahle, M. Schreckenberg: %, 2000,
\tit{Microscopic simulation of evacuation processes on passenger ships}
In: {\em Theoretical and Practical Issues on Cellular Automata}
%{\em Proceedings of the 4th International Conference on CA for Research and
%Industry},
ed. by S. Bandini, T. Worsch
(Springer, London 2000) \pag{63}{71}
\bib{HelFaVi00b}
D. Helbing, I. J. Farkas, T. Vicsek: %, 2000b,
\tit{Simulating dynamical features of escape panic}
\art{Nature}{407}{487}{490}{(2000)}
\bib{Fasold} D. Fasold: %, 2001,
\tit{Optimierung logistischer Prozessketten am Beispiel einer Nass\"atzanlage in der
Halbleiterproduktion}
Master's thesis, TU Dresden (2001)
\end{thebibliography}
\end{document}
\bib{zzz} ZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZZ
\bibitem{journ1} W. Frank, A. Seeger: Appl. Phys. A \textbf{3}, 66 (1988)
\bibitem{proc1} W. Greiner, D.N. Poenaru: `Cluster Preformation in
Closed- and Mid-shell Nuclei'. In:
\emph{Atomic and Nuclear Clusters, 2nd International Conference at Santorini, Greece,
June 28--July 2, 1993}, ed. by G.S. Anagnostatos, W. von Oertzen (Springer, Heidelberg 1994)
pp. 264--266
\bibitem{fhjl1} F. Holzwarth, J. Lenz et al.: \emph{1readme.
Further Details on Layout and \LaTeX\, code}. (Springer, Berlin
Heidelberg 1999)
\bibitem{mono1} B. Jirgensons: \emph{Optical Activity for Proteins
and Other Macro-Molecules}, 2nd edn.
(Springer, New York 1984)
\bibitem{contr1} D.M. MacKay: `Visual Stability and Voluntary Eye
Movements'. In: \emph{Handbook of
Sensory Physiology VII/3}. ed. by R. Jung (Springer, Berlin, Heidelberg 1973) pp. 307--331
\bibitem{mono2} M. M\"uller, F.J. Becker: \emph{On Generalized
Hamiltonian Dynamics} (Cambridge University Press, Cambridge 1930)
\bibitem{journ2} S. Nakamura, M. Senoh, N. Iwasa, S. Nagahama: Jpn. J.
Appl. Phys. \textbf{34}, L797 (1995)
W. Frank, A. Seeger: Appl. Phys. A \textbf{3}, 66 (1988)
\bibitem{thesis1} D.W. Ross: Lysosomes and Storage Diseases. MA Thesis,
Columbia University, New York
(1977)
\end{thebibliography}
%INDEX%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Please check with the editor of your book whether he plans to
% include a "mutual" subject index - if so, please code your entries
% in the standard syntax. For your own purposes you may print your
% "personal" index by using the following commands:
%
%\clearpage
%\addcontentsline{toc}{section}{Index}
%\flushbottom
%\printindex
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\end{document}