1: \documentclass[twoside]{article}
2: \usepackage{docbib}
3: \usepackage{graphicx}
4: %\usepackage{fancyheadings}
5: %\usepackage{emlines}
6:
7: \textheight180mm %200mm
8: \textwidth140mm
9: \columnsep6mm
10: \parsep-10mm
11: %%%%%%%%%%%%\oddsidemargin-10mm
12: %%%%%%%%%%%%%%\topmargin-20mm
13: \sloppy
14: \topmargin-0.5cm
15: \oddsidemargin0.2cm
16: \evensidemargin0.2cm
17: \frenchspacing
18: %\setlength{\parindent}{0pt}
19: %\setlength{\parskip}{10pt plus 4pt minus 2pt}
20:
21: \newfont{\sss}{cmbxti10} % scaled 1100}
22: \newfont{\kleiner}{cmbx9}
23: \newfont{\foot}{cmr9}
24: \newfont{\ital}{cmti9}
25: \renewcommand{\it}{\ital}
26: \renewcommand{\footnotesize}{\foot}
27: \renewcommand{\small}{\foot}
28: \newfont{\Gross}{cmbx10 scaled 1400}
29: \renewcommand{\Large}{\large}
30: \renewcommand{\thesection}{\arabic{section}.\hspace*{-0.5em}}
31: \renewcommand{\thesubsection}{\thesection.\arabic{subsection}.\normalsize}
32: \renewcommand{\thesubsubsection}{\normalsize\hspace*{-1em}}
33:
34: \setcounter{topnumber}{4}
35: \renewcommand\topfraction{1.0} %{.7}
36: \setcounter{bottomnumber}{4}
37: \renewcommand\bottomfraction{0.0} %{.3}
38: \setcounter{totalnumber}{4}
39: \renewcommand\textfraction{.0}
40: \renewcommand\floatpagefraction{1.0}
41: \setcounter{dbltopnumber}{4}
42: \renewcommand\dbltopfraction{1.0} %{.7}
43: \renewcommand\dblfloatpagefraction{1.0}
44:
45: \newcommand{\abb}[3]{\begin{figure}[h]
46: \begin{center}\vspace*{-2mm} #1 \end{center} \vspace*{-3mm}
47: {\hfill\parbox[b]{5.5cm}{\caption[]{#2\label{#3}}}\hfill}
48: \end{figure}}
49:
50: \renewcommand{\vec}[1]{\relax\ifmmode\mathchoice
51: {\mbox{\boldmath$\relax\displaystyle#1$}}
52: {\mbox{\boldmath$\relax\textstyle#1$}}
53: {\mbox{\boldmath$\relax\scriptstyle#1$}}
54: {\mbox{\boldmath$\relax\scriptscriptstyle#1$}}\else
55: \hbox{\boldmath$\relax\textstyle#1$}\fi}
56:
57: \newcommand{\abbl}[4]{%
58: \begin{figure}[htbp]\vspace*{0.5cm}%
59: \parbox[b]{6cm}{\begin{center} #1 \end{center}}#4\hfill%
60: \parbox[b]{9cm}{%
61: \caption[#2]{#3\\ \mbox{ }}
62: }%
63: \vspace*{0.5cm}\end{figure}}
64: \unitlength=1cm
65:
66: \begin{document}
67: \thispagestyle{empty}
68: \pagestyle{myheadings}
69: \markboth{{\kleiner INTERNATIONAL JOURNAL OF CHAOS THEORY AND
70: APPLICATIONS\mbox{\ \ \ }}}
71: {\mbox{\ \ \ \ \ }
72: {\em Dirk Helbing and Tadeusz Platkowski}{\rm : Self-Organization
73: Induced by Fluctuations}}
74:
75: \onecolumn
76: \twocolumn[\protect\hsize\textwidth\columnwidth\hsize\csname @twocolumnfalse\endcsname
77: {\vspace*{-15mm}
78: \begin{center}
79: {\bf INTERNATIONAL JOURNAL OF CHAOS THEORY AND APPLICATIONS\\
80: Volume ? (2000), Nr. ? }\\
81: \mbox{}\hrulefill\mbox{}\\[9mm]
82: \,\hfill {\sss Invited Paper}\\[9mm]
83: {\Gross Self-Organization in Space and Induced by Fluctuations}\\[1cm]
84: {\large
85: Dirk Helbing$^{\circ\,\ast}$ and Tadeusz Platkowski$^{\circ\,+}$}\\[1mm]
86: %\address{
87: %\normalsize {\tt http://www.theo2.physik.uni-stuttgart.de/helbing.html}\\[-1mm]
88: %\normalsize {\tt http://hydra.mimuw.edu.pl/$\widetilde{\hphantom{I}}$tplatk/ }}
89: {\normalsize\it $^\circ$ II. Institute of Theoretical Physics, University of Stuttgart,\\
90: \normalsize\it Pfaffenwaldring 57/III, 70550 Stuttgart, Germany\\
91: \normalsize\it $^\ast$ Collegium Budapest, Institute of Advanced Study,
92: Szeth\'aroms\'ag u. 2, \\
93: \normalsize\it H-1014 Budapest, Hungary\\
94: \normalsize\it $^+$Dept. of Mathematics, Informatics and Mechanics,\\
95: \normalsize\it University of Warsaw, Banacha 2, 02-097 Warsaw, Poland
96: }\\[6mm]
97: %{\tt helbing@theo2.physik.uni-stuttgart.de; tplatk@mimuw.edu.pl}}
98: \end{center}\par
99: {\bf Abstract\footnotemark[1]}\\[1mm]
100: We present a simple discrete model for the non-linear spatial interaction of
101: different kinds of ``subpopulations'' composed of identical moving entities
102: like particles, bacteria, individuals, etc. The model allows to mimic a variety
103: of self-organized agglomeration and segregation phenomena. By relating it to
104: game-theoretical ideas, it can be applied not only to attractive and
105: repulsive interactions in physical and chemical systems, but also to the
106: much richer combinations of positive and negative interactions found in
107: biological and socio-economic systems. Apart from investigating symmetric
108: interactions related to a continuous increase of the ``overall success''
109: within the system (``self-optimization''), we will focus on cases, where
110: fluctuations further or induce self-organization, even though the initial
111: conditions and the interactions are assumed homogeneous in space
112: (translation invariant).\\[5mm]
113: {\bf Keywords:} {\sss Self-organization, self-optimization, game theory,
114: fluctuation-induced transition, agglomeration, segregation.}\\[9mm]\mbox{ }
115: }
116: ]
117: \footnotetext[1]{Manuscript received: xxxx; accepted: xxxx}
118:
119: \section{\large Introduction}
120:
121: Although the biological, social, and economic world are full of
122: self-organization phenomena, many people believe that the dynamics behind them
123: is too complex to be modelled in a mathematical way.
124: Reasons for this are the huge
125: number of interacting variables, most of which cannot be quantified, plus
126: the assumed freedom of decision-making or large fluctuations within
127: biological and socio-economic systems. However, in many situations, the
128: living entities making up these systems decide for some (more or less)
129: optimal behavior, which can make the latter describable or predictable to a
130: certain extend
131: \cite{We91,game1,game2,game3,gamedyn,Hub,quantsoc,frank,Lewenstein,Galam}.
132: %(think of prognoses of the rough outcome of elections).
133: This is even more the case for the behavior
134: shown under certain constraints like, for
135: example, in pedestrian or vehicle
136: dynamics \cite{book,pre,phasediag}. While
137: pedestrians or vehicles can move freely at small traffic densities, at large
138: densities the interactions with the others and with the boundaries of the
139: street confines them to a small spectrum of moving behaviors. Consequently,
140: empirical traffic dynamics can be reproduced by simulation models
141: surprisingly well \cite{book,pre,phasediag,solid,scatter,opus,trail}.
142: \par
143: In this connection, it is also interesting to mention some insights gained
144: in statistical physics and complex systems theory: Non-linearly interacting
145: variables do not change independently of each other, and in many cases there
146: is a separation of the time scales on which they evolve. This often allows
147: to ``forget'' about the vast number of rapidly changing variables, which are
148: usually determined by a small number of ``order parameters'' and treatable as
149: fluctuations \cite{Haken,Adv}. In the above mentioned
150: examples of traffic dynamics, the order parameters are the traffic density
151: and the average velocity of pedestrians or vehicles.
152: \par
153: Another discovery is that, by proper transformations or scaling, many
154: different models can by mapped onto each other, i.e.
155: they behave basically the same \cite{Haken,Adv,seg3,phasediag}.
156: That is, a certain class of models displays the same kinds of
157: states, shows the same kinds of transitions among them, and can be described
158: by the same ``phase diagram'', displaying the respective states as a function
159: of some ``control parameters'' \cite{Haken,Adv}.
160: We call such a class of models a
161: ``universality class'', since any of these models shows the same kind of
162: ``universal'' behavior, i.e., the same phenomena. Consequently, one usually tries
163: to find the simplest model having the properties of the universality class.
164: While physicists like to call it a ``minimal model'', ``prototype model'', or
165: ``toy model'', mathematicians named the corresponding mathematical equations
166: ``normal forms'' \cite{manneville,Zee77,Haken,Adv}.
167: \par
168: Universal behavior is the reason of the great
169: success of systems theory \cite{Ber68,Bu67,Rap86}
170: in comparing phenomena in seemingly
171: completely different systems, like physical, biological, or social ones.
172: However, since these systems are composed of different entities and their
173: corresponding interactions can be considerably different, it is not always
174: easy to identify the variables and parameters behind their dynamics. It can
175: be helpful to take up game-theoretical ideas, here, quantifying
176: interactions in terms of payoffs
177: \cite{game1,game2,game3,gamedyn,Hub,quantsoc,Ebeling,gd}.
178: This can be applied to
179: positive (profitable, constructive, cooperative, symbiotic) or
180: negative (competitive, destructive) interactions in socio-economic or
181: biological systems, but to attractive and repulsive interactions in
182: physical systems as well
183: \cite{helvic}.
184: %Info,Pri,Haken,evol,Ebeling,gamedyn,quantsoc,frank,eshel,biol1a,biol1b,%
185: %biol1c,biol2,biol3,biol4,biol5,biol6,Drasdo,ants1,ants2,We91,Hub,Hel93,Galam,
186: %pre,granular,solid,trail,helvic}.
187: \par
188: In the following, we will investigate a simple model for interactive motion
189: in space allowing to describe (i) various self-organized agglomeration
190: phenomena, like settlement formation, and segregation phenomena, like ghetto
191: formation, emerging from different kinds of interactions and (ii)
192: fluctuation-induced ordering or self-organization phenomena.
193: \par
194: Noise-related phenomena can be quite surprising and have, therefore,
195: recently attracted the interest of many researchers. For example, we
196: mention stochastic resonance \cite{stochres},
197: noise-driven motion \cite{biol1a,biol1c}, and ``freezing by heating''
198: \cite{freezing}.
199: \par
200: The issue of order through fluctuations
201: has already a considerable history. Prigogine has discussed it
202: in the context of structural instability with respect to the
203: appearance of a new species \cite{Pri,Pri2}, but this is not related
204: to the approach considered in the following.
205: \par
206: Moreover, since both, the
207: initial conditions and the interaction strengths in our model are
208: assumed independent of the position in space,
209: the fluctuation-induced self-organization
210: discussed later on must be distinguished from
211: so-called ``noise-induced transitions'' as well,
212: where we have a space-dependent diffusion coefficient which
213: can induce a transition \cite{HoLe84}.
214: \par
215: Although our model
216: is related to diffusive processes,
217: it is also different from reaction-diffusion systems
218: that can show fluctuation-induced self-organization phenomena known as
219: Turing patterns
220: \cite{turing,turing1,turing2,turing3,turing4,turing5},
221: which are usually periodic in space. The
222: noise-induced self-organization that we find seems to have (i) no typical
223: length scale and (ii) no attractor, since our model is
224: translation-invariant.
225: This, however,
226: is not yet a final conclusion and still subject to investigations.
227: \par
228: We also point out that, in the case of spatial invariance, self-organization
229: directly implies spontaneous symmetry-breaking, and we expect a pronounced
230: history-dependence of the resulting state. Nevertheless, when averaging over a
231: large ensemble of simulation runs with different random seeds, we again
232: expect a homogeneous distribution, since this is the only result compatible
233: with translation invariance.
234: \par
235: Finally, we mention that our results do not fit into the concept of
236: %are not consistent with
237: noise-induced transitions from a metastable disordered
238: state (local optimum) to a stable ordered
239: state (global optimum),
240: which are, for example, found for undercooled fluids, metallic glasses,
241: or some granular systems \cite{gran1,gran2,gran3}.
242:
243: \section{\large Discrete Model of Interactive Motion in Space}
244: %\vspace*{4mm}
245:
246: Describing motion in space has the advantage that the essential variables
247: like positions, densities, and velocities are well measurable, which allows
248: to calibrate, test, and verify or falsify the model. Although we will focus
249: on motion in ``real'' space like the motion of pedestrians or bacteria, our
250: model may also be applied to changes of positions in abstract spaces, e.g. to
251: opinion changes on an opinion scale \cite{quantsoc,Hel93}. %,Schw91}.
252: There exist, of course,
253: already plenty of models for motion in space, and we can mention only a few
254: \cite{We91,quantsoc,book,pre,phasediag,solid,scatter,opus,trail,Haken,Adv,seg3,%
255: manneville,Ebeling,helvic,biol1a,biol1c,freezing,%
256: HoLe84,turing,turing1,turing2,turing3,turing4,%
257: turing5,gran2,Keizer,Hel93,granular,eshel,jacob2,biol6,biol2,%
258: biosys,millonas}. Most of them are,
259: however, rather specific for certain systems, e.g., for fluids or for
260: migration behavior.
261: \par
262: For simplicity, we will restrict the following considerations to a
263: one-dimensional space, but a generalization to higher dimensions is
264: straightforward. The space is divided into $I$ equal cells $i$ which can be
265: occupied by the entities. We will apply periodic boundary conditions, i.e.
266: the space can be imagined as a circle. In our model, we group the $N$
267: entities $\alpha$ in the system into homogeneously behaving subpopulations $%
268: a $. If $n_i^a(t)$ denotes the number of entities of subpopulation $a$ in
269: cell $i$ at time $t$, we have the relations
270: \begin{equation}
271: \sum_i n_i^a(t) = N_a, \qquad
272: \sum_a N_a = N.
273: \end{equation}
274: We will assume that the numbers $N_a$ of entities belonging to the
275: subpopulations $a$ do not change. It is, however, easy to take additional
276: birth and death processes and/or transitions of individuals from one
277: subpopulation to another into account \cite{We91}.
278: \par
279: In order not to introduce any bias, we start our simulations with a
280: completely uniform distribution of the entities in each subpopulation over
281: the $I$ cells of the system, i.e., $n_i^a(0) = n_{\rm hom}^a
282: = N_a/I$, for which we choose a
283: natural number. At times $t \in \{1,2,3,...\}$, we apply the following
284: update steps, using a random sequential update (although a parallel
285: update is possible as well,
286: which is more efficient \cite{Wolfram,Stauffer}, but normally less realistic
287: \cite{Bernardo} due to the assumed synchronous updating):
288: \par
289: {\em 1st step:\/} For updating of the state of entity $\alpha$, given
290: it is a member of subpopulation $a$ and located in cell $i$,
291: determine the so-called (expected) ``success'' according to the formula
292: \begin{equation}
293: S_{a}(i,t)=\sum_{b}P_{ab}\, n_{i}^{b}(t)+\xi _{\alpha }(t) \, .
294: \label{formu}
295: \end{equation}
296: Here, $P_{ab}$ is the ``payoff'' in interactions of an entity of
297: subpopulation $a$ with an entity of subpopulation $b$.
298: The payoff $P_{ab}$ is positive for attractive, profitable, constructive, or
299: symbiotic interactions, while it is negative for repulsive, competitive, or
300: destructive interactions. Notice that $P_{ab}$ is assumed
301: to be independent of the position (i.e., translation invariant), while the
302: total payoff $\sum_{b}P_{ab}\,n_{i}^{b}(t)$ due to interactions
303: depends on the distribution of entities over the system.
304: The latter is an essential point for the possibility of
305: fluctuation-induced self-organization. We also point out that,
306: in formula (\ref{formu}), pair interactions are restricted to
307: the cell in which the
308: individual is located. Therefore, we do not assume spin-like or Ising-like
309: interactions, %like, e.g.,
310: %\begin{equation}
311: %S'_{a}(i,t)= \sum_{b}\sum_{j\in\{i\pm 1\}} P_{ab}\, n_{i}^{a}(t)n_j^b(t)
312: %+\xi _{\alpha }(t)
313: %\end{equation}
314: %would. This is
315: in contrast to other
316: quantitative models proposed for the
317: description of social behavior \cite{Lewenstein,Galam}.
318: \par
319: The quantities $\xi_\alpha(t)$ are random
320: variables allowing to consider individual
321: variations of the success, which may be ``real'' or due to uncertainty
322: in the evaluation or estimation of success.
323: In our simulation program, they are uniformly
324: distributed in the interval $[0,D_a]$, where $D_a$ is the
325: fluctuation strength (not to be mixed up wich a diffusion constant).
326: However, other specifications of the noise term are possible as well.
327: \par
328: {\em 2nd step:\/} Determine the (expected) successes $S_a(i\pm 1,t)$
329: for the nearest neighbors $(i\pm 1)$ as well.
330: \par
331: {\em 3rd step:\/} Keep entity $\alpha$ in its previous cell $i$,
332: if $S_a(i,t) \ge \max \{S(i-1,t),S(i+1,t)\}$. Otherwise, move to cell
333: $(i-1)$, if $S(i-1,t) > S(i+1,t)$, and move to cell $(i+1)$,
334: if $S(i-1,t) < S(i+1,t)$. In the remaining case $S(i-1,t) = S(i+1,t)$,
335: jump randomly to cell $(i-1)$ or $(i+1)$ with probability 1/2.
336: \par
337: If there is a maximum density $\rho_{\rm max} = N_{\rm max}/I$
338: of entities, overcrowding can be avoided by introducing a
339: saturation factor
340: \begin{equation}
341: c(j,t) = 1 - \frac{N_j(t)}{N_{\rm max}}\, , \quad
342: N_j(t) = \sum_a n_j^a(t) \, ,
343: \end{equation}
344: and performing the update steps with the generalized success
345: \begin{equation}
346: S'_a(j,t) = c(j,t) S_a(j,t)
347: %\bigg[ \sum_b P_{ab} \, n_j^b(t) + \xi_\alpha(t) \bigg]
348: \end{equation}
349: instead of $S_a(j,t)$, where $j \in \{i-1,i,i+1\}$. The model can be also
350: easily extended to include long distance interactions,
351: jumps to more remote cells, etc. (cf. Section 5).
352:
353: \section{\large Simulation Results}
354:
355: We consider two subpopulations $a\in\{1,2\}$ and $N_1 = N_2 = 100$
356: entities in each subpopulation, which are distributed over $I=20$ cells.
357: The payoff matrix $(P_{ab})$ will be represented by the vector
358: $\vec{P} = (P_{11}, P_{12}, P_{21}, P_{22})$,
359: where we will restrict ourselves
360: to $|P_{ab}|\in\{ 1, 2\}$ for didactical reasons.
361: For symmetric interactions between subpopulations, we have
362: $P_{ab} = P_{ba}$, while for
363: asymmetric interactions, there is $P_{ab} \ne P_{ba}$, if $a\ne b$.
364: For brevity, the interactions within the same population will be called
365: self-interactions, those between different populations cross-interactions.
366: \par
367: To characterize the level of self-organization in each subpopulation
368: $a$, we can, for example, use the
369: overall successes
370: \begin{equation}
371: S_a(t) = \frac{1}{I^2} \sum_i \sum_b n_i^a(t) \, P_{ab} \, n_i^b(t) \, ,
372: \end{equation}
373: %\begin{equation}
374: %S(t) = \sum_{a=1}^2 S_a(t) \, , \ \
375: %\end{equation}
376: the variances
377: \begin{equation}
378: V_a(t) = \frac{1} {I^2} \sum_i \, [n^a_i(t)-n^a_{\rm hom}]^2 \, ,
379: \end{equation}
380: %\begin{equation}
381: %V(t) = \frac{1} {I^2} \sum_{j=1}^I \, [n_j(t)- n_{\rm hom}]^2 \, , \ \
382: %\end{equation}
383: or the alternation strengths
384: \begin{equation}
385: A_a(t) = \frac{1} {I^2} \sum_i \, [n^a_i(t)-n^a_{i-1}(t)]^2 \, .
386: \end{equation}
387: %\begin{equation}
388: %A(t) = \frac{1} {I^2} \sum_{j=1}^I \, [n_j(t)-n_{j-1}(t)]^2 \, ,
389: %\end{equation}
390: %where
391: %\begin{equation}
392: %n_{\rm hom} = n^1_{\rm hom} + n^2_{\rm hom}, \ \ n^a_{\rm hom}=N_a/I.
393: %\end{equation}
394:
395: \subsection*{\normalsize 3.1. Symmetric Interactions}
396:
397: By analogy with a more complicated model \cite{helvic} it is expected
398: that the global overall success $S(t)=\sum_a S_a(t)$
399: is an increasing function in time, if the
400: fluctuation strengths $D_a$ are zero. However, what happens at finite noise
401: amplitudes $D_a$ is not exactly known. One would usually expect that finite
402: noise tends to obstruct or suppress self-organization, which will
403: be investigated in the following.
404: \par
405: We start with the payoff matrix $\vec{P}=(2,-1,-1,2)$ corresponding to
406: positive (or attractive) self-interactions and negative
407: (or repulsive) cross-interactions. That is, entities of the same
408: subpopulation like each other, while entities of different
409: subpopulations dislike each other. The result will naturally be
410: segregation (``ghetto formation'') \cite{Sche71,We91},
411: if the noise amplitude is small.
412: However, segregation is suppressed by large fluctuations,
413: as expected (see Fig.~\ref{fig1}).
414: \par
415: \abb{\hspace*{-2mm}\includegraphics[height=5.5cm, angle=-90]{fig1a.ps}\\[-7.5mm]
416: \hspace*{-2mm}\includegraphics[height=5.5cm, angle=-90]{fig1b.ps}}
417: {Resulting distribution of entities at $t=4000$ for
418: the payoff matrix $\vec{P}=(2,-1,-1,2)$ at small fluctuation strength
419: $D_a=0.1$ (top) and large fluctuations strength $D_a=5$ (bottom).}
420: {fig1}
421: However, for medium noise amplitudes $D_a$, we find a much more pronounced
422: self-organization (segregation) than for small ones (compare
423: Fig.~\ref{fig2} with Fig.~\ref{fig1}). The effect is systematic
424: insofar as the degree of segregation (and, hence, the overall success)
425: increases with increasing noise amplitude, until segregation breaks
426: down above a certain critical noise level.
427: %Note that the increasing
428: %level of ordering is not so easy to understand, since
429: %in the one-dimensional case, the repulsive
430: %clusters of different subpopulations cannot simply pass each other in
431: %order to join others of the same subpopulation.
432: \par\abb{\includegraphics[height=6.0cm, angle=-90]{fig2a.ps}\\[-7.5mm]
433: \includegraphics[height=6.0cm, angle=-90]{fig2b.ps}}
434: {As Fig.~\ref{fig1}, but with medium fluctuation strength
435: $D_a=2$ (top) and $D_a=3$ (bottom).}
436: {fig2}
437: Let us investigate some other cases: For the structurally similar
438: payoff matrix $(1,-2,-2,1)$,
439: we find segregation as well, which is not surprising. In contrast, we
440: find agglomeration for the payoff matrices
441: $(1,2,2,1)$ and $(2,1,1,2)$. This agrees with intuition, since all
442: entities like each other in these cases, which makes them move to
443: the same places, like in the formation of settlements \cite{We91}, the
444: development of trail systems \cite{trail,biosys,millonas}, or the build up of
445: slime molds \cite{biol6,Pri2}. More interesting is the case corresponding to
446: the payoff matrix $(-1,2,2,-1)$, where the cross-interactions are positive
447: (attractive), while the self-interactions are
448: negative (repulsive). One may think that this causes the entities of the same
449: subpopulation to spread homogeneously over the system,
450: and in all cells would result an equal number of entities
451: of both subpopulations, which is compatible with mutual attraction.
452: However, this homogeneous distribution turns out to be unstable with
453: respect to fluctuations. Instead, we find agglomeration! This result
454: is more intuitive if we imagine one subpopulation to represent
455: women and the other one men (without taking this example too serious).
456: While the interaction between women
457: and men is normally strongly attractive, the interactions among men or
458: among women may be considered to be weakly competitive. As we all
459: know, the result is a tendency of young men and women to move into
460: cities. Corresponding simulation results for different noise strengths
461: are depicted in Fig.~\ref{fig3}. Again, we find that the
462: self-organized pattern is destroyed by strong fluctuations
463: in favour of a more or less
464: homogeneous distribution, while medium noise strengths
465: further self-organization.
466: \par\abb{
467: \hspace*{-2mm}\includegraphics[height=5.5cm, angle=-90]{fig3a.ps}\\[-7.5mm]
468: \includegraphics[height=6.0cm, angle=-90]{fig3b.ps}\\[-7.5mm]
469: \hspace*{-2mm}\includegraphics[height=5.5cm, angle=-90]{fig3c.ps}}
470: {As Fig.~\ref{fig1}, but for the payoff matrix
471: $\vec{P}=(-1,2,2,-1)$ and
472: $D_a=0.05$ (top), $D_a=1.5$ (middle), and $D_a=5$ (bottom).}
473: {fig3}
474: For the payoff matrices $(-2,1,1,-2)$ and $(-2,-1,-1,-2)$, i.e. cases
475: of strong negative self-interactions, we find
476: a more or less homogeneous distribution of entities in both
477: subpopulations, irrespective of the noise amplitude. In contrast,
478: the payoff matrix $(-1,-2,-2,-1)$ corresponding to negative
479: self-interactions but even stronger negative cross-interactions,
480: leads to another self-organized pattern. We may describe it
481: as the formation of lanes, as it is observed in pedestrian counterflows
482: \cite{pre,helvic} or in sheared granular media with different
483: kinds of grains \cite{granular}.
484: While both subpopulations tend to separate from
485: each other, at the same time they tend to spread over all the
486: available space (see Fig.~\ref{fig4}),
487: in contrast to the situation depicted in Figs.~\ref{fig1} and
488: \ref{fig2}. Astonishingly enough, a medium level of noise again
489: supports self-organized ordering, since it
490: helps the subpopulations to separate from each other.
491: \par\abb{
492: \hspace*{-2mm}\includegraphics[height=5.5cm, angle=-90]{fig4a.ps}\\[-7.5mm]
493: \hspace*{-2mm}\includegraphics[height=5.5cm, angle=-90]{fig4b.ps}\\[-7.5mm]
494: \hspace*{-2mm}\includegraphics[height=5.5cm, angle=-90]{fig4c.ps}}
495: {As Fig.~\ref{fig1}, but for the payoff matrix
496: $\vec{P}=(-1,-2,-2,-1)$ and
497: $D_a=0.05$ (top), $D_a=0.5$ (middle), and $D_a=5$ (bottom).}
498: {fig4}
499: We finally mention that a
500: finite saturation level suppresses self-organization in a
501: surprisingly strong way, as is shown in Fig.~\ref{fig5}. Instead of
502: pronounced segregation, we will find a result similar to lane
503: formation, and even strong agglomeration will be replaced by
504: an almost homogeneous distribution.
505: \par\abb{\hspace*{-2mm}\includegraphics[height=5.5cm, angle=-90]{fig5a.ps}\\[-7.5mm]
506: \hspace*{-2mm}\includegraphics[height=5.5cm, angle=-90]{fig5b.ps}}
507: {Resulting distribution of entities at $t=4000$ with saturation level
508: $N_{\rm max}=50$. Top: $\vec{P}=(2,-1,-1,2)$ and $D_a=3$.
509: Bottom: $\vec{P}=(-1,2,2,-1)$ and $D_a=1.5$.}
510: {fig5}
511:
512: \subsubsection*{{\sss Noise-induced ordering}}
513: %xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
514: A possible interpretation for noise-induced ordering would be that
515: fluctuations allow the system to leave local minima (corresponding to
516: partial agglomeration or segregation only). This could trigger
517: a transition to a more stable state with more pronounced
518: ordering. However, although this interpretation is consistent
519: with a related example discussed in Ref.~\cite{helvic}, the idea of
520: a step-wise coarsening process is not supported by the
521: %In order to understand the fact that a medium level of fluctuation
522: %furthers self-organization in our model, let us study the
523: temporal evolution of the distribution of entities (see Fig.~\ref{fig6}) and
524: the time-dependence of the overall success within the subpopulations
525: (see Fig.~\ref{fig7}). This idea %of a coarsening dynamics
526: is anyway not applicable to segregation, since,
527: in the one-dimensional case, the repulsive
528: clusters of different subpopulations cannot simply pass each other in
529: order to join others of the same subpopulation.
530: \par
531: According to Figs.~\ref{fig6} and \ref{fig7},
532: segregation and agglomeration rather take place in
533: three phases: First, there is a certain time interval, during which
534: the distribution of entities remains more or less homogeneous. Second,
535: there is a short period of rapid self-organization. Third, there is
536: a continuing period, during which the distribution and overall success
537: do not change anymore. The latter is a consequence of the short-range
538: interactions within our model, which are limited to the nearest
539: neighbors. Therefore, the segregation or aggregation process
540: practically stops, after separate peaks have evolved. This is not the
541: case for lane formation, where the entities redistribute, but
542: all cells remain occupied, so that we have ongoing interactions. This
543: is reflected in the non-stationarity of the lanes and by the
544: oscillations of the overall success.
545: \par\abb{
546: \vspace*{-5mm}
547: \hspace*{6mm}\includegraphics[height=6cm, angle=-90]{fig6a.ps}\\[-10mm]
548: \hspace*{6mm}\includegraphics[height=6cm, angle=-90]{fig6b.ps}\\[-10mm]
549: \hspace*{6mm}\includegraphics[height=6cm, angle=-90]{fig6c.ps}
550: \vspace*{-5mm}}
551: {Temporal evolution of the distribution of entitities within subpopulation
552: $a=2$ for $\vec{P}=(2,-1,-1,2)$ and $D_a=3$ (top),
553: $\vec{P}=(-1,2,2,-1)$ and $D_a=1.5$ (middle), and
554: $\vec{P}=(-1,-2,-2,-1)$ and $D_a=0.5$ (bottom).}
555: {fig6}
556: \par\abb{
557: \hspace*{-2mm}\includegraphics[height=5.5cm, angle=-90]{fig7a.ps}\\[-7.5mm]
558: \hspace*{-2mm}\includegraphics[height=5.5cm, angle=-90]{fig7b.ps}\\[-7.5mm]
559: \hspace*{-2mm}\includegraphics[height=5.55cm, angle=-90]{fig7c.ps}}
560: {Temporal evolution of the overall success within both subpopulations
561: for $\vec{P}=(2,-1,-1,2)$ and $D_a=3$ (top),
562: $\vec{P}=(-1,2,2,-1)$ and $D_a=1.5$ (middle), and
563: $\vec{P}=(-1,-2,-2,-1)$ and $D_a=0.5$ (bottom).}
564: {fig7}
565: %For segregation, it is particularly hard to imagine how larger noise
566: %brings about more pronounced segregation, since
567: %in the one-dimensional case, the repulsive
568: %%clusters of different subpopulations cannot simply pass each other in
569: %order to join others of the same subpopulation. Anyway, the idea of
570: %an eventual coarsening process is not really supported by
571: %the simulation results depicted in Fig.~\ref{fig5}, in contrast to
572: %a slightly different example discussed in Ref.~\ref{helvic}.
573: We suggest the following interpretation for the three phases mentioned
574: above: %It rather appears that a long-range pre-ordering process takes place
575: During the first time interval, which is
576: characterized by a quasi-continous distribution of entities over
577: space, a long-range pre-ordering process takes place.
578: After this ``phase of preparation'',
579: order develops in the second phase similar to crystallization, and
580: it persists in the third phase. The role of fluctuations seems to be the
581: following: An increased noise level avoids a rash local
582: self-organization by keeping up a quasi-continuous distribution of
583: entities, which is required for a redistribution of entities over
584: larger distances. In this way, a higher noise level increases the effective
585: interaction range by extending the first phase, the ``interaction phase''.
586: As a consequence, the resulting structures are more extended in space
587: (but probably without a characteristic length scale, see
588: Introduction).
589: \par
590: It would be interesting to investigate, whether this
591: mechanism has something to do with the recently discovered phenomenon
592: of ``freezing by heating'', where a medium noise level causes
593: a transition to a highly ordered (but energetically less stable)
594: state, while extreme noise levels
595: produce a disordered, homogeneous state again \cite{freezing}.
596:
597: \subsection*{\normalsize 3.2. Asymmetric Interactions}
598:
599: Even more intriguing transitions than in the symmetric case can be
600: found for %the numerous cases of
601: asymmetric interactions between the subpopulations. Here,
602: we will focus on the payoff matrix $(-1,2,-2,1)$, only. This example
603: corresponds to the curious case, where individuals of subpopulation 1
604: weakly dislike each other, but strongly like individuals of the
605: other subpopulation. In contrast, individuals of subpopulation 2
606: weakly like each other, but they strongly dislike the other
607: subpopulation. A good example for this is hard to find. With some good
608: will, one may imagine subpopulation 1 to represent poor
609: people, while subpopulation 2 corresponds to rich people. What will be
610: the outcome? In simple terms, the rich are expected to agglomerate in
611: a few areas, if the poor are moving too nervously (see
612: Fig.~\ref{fig8}). In detail,
613: however, the situation is quite complex, as discussed in the next
614: paragraph.
615: \abb{
616: \hspace*{-2mm}\includegraphics[height=5.5cm, angle=-90]{fig8a.ps}\\[-7.5mm]
617: \hspace*{-2mm}\includegraphics[height=5.5cm, angle=-90]{fig8b.ps}\\[-7.5mm]
618: \includegraphics[height=6.0cm, angle=-90]{fig8c.ps}\\[-7.5mm]
619: \hspace*{-2mm}\includegraphics[height=5.5cm, angle=-90]{fig8d.ps}}
620: {Distributions for %As Fig.~\ref{fig1}, but for the payoff matrix
621: $\vec{P}=(-1,2,-2,1)$ and
622: $D_1=D_2=0.5$ (top), $D_1=50$, $D_2=0.5$ (second),
623: $D_1=5000$, $D_2=0.5$ (third), $D_1=5000$, $D_2=5$ (bottom).}
624: {fig8}
625:
626: \subsubsection*{\sss Noise-induced self-organization}
627:
628: At small noise levels $D_a$,
629: we will just find more or less
630: homogeneous distributions of the entities. This
631: is already different from the cases of agglomeration,
632: segregation, and lane formation we have
633: discussed before. Self-organization is also not found at higher
634: noise amplitudes $D_a$,
635: as long as we assume that they are the same in both subpopulations
636: (i.e., $D_1 = D_2$).
637: However, given that the fluctuation amplitude $D_2$ in subpopulation 2 is
638: small, we find an agglomeration in subpopulation 2, if the noise level
639: $D_1$ in subpopulation 1 is medium or high, so that subpopulation 1
640: remains homogeneously distributed. The order in subpopulation
641: 2 breaks down, as
642: soon as we have a relevant (but still small) noise level $D_2$ in
643: subpopulation 2 (see Fig.~\ref{fig8}).
644: \par
645: Hence, we have a situation
646: where asymmetric noise with $D_1\ne D_2$ can facilitate self-organization in a
647: system with completely homogeneous initial conditions and interaction
648: laws, where we would not have ordering without any noise. We call this
649: phenomenon noise-induced self-organization. It is to be distinguished
650: from the noise-induced increase in the degree of ordering discussed
651: above, where we have self-organization even without noise,
652: if only the initial conditions are not fully homogeneous.
653: \par
654: The role of the noise in subpopulation 1 seems to be the following: Despite of
655: the attractive interaction with subpopulation 2, it
656: suppresses an agglomeration in subpopulation 1,
657: in particular at the places where subpopulation 2 agglomerates.
658: Therefore, the repulsive interaction of
659: subpopulation 2 with subpopulation 1 is effectively reduced. As a
660: consequence, the attractive self-interaction within subpopulation 2
661: dom-\linebreak\newpage
662: \noindent inates, which gives rise to the observed agglomeration.
663: %\par\abb{
664: %\hspace*{-2mm}\includegraphics[height=5.5cm, angle=-90]{fig8a.eps}\\[-7.5mm]
665: %\includegraphics[height=6.0cm, angle=-90]{fig8b.eps}\\[-7.5mm]
666: %\hspace*{-2mm}\includegraphics[height=5.5cm, angle=-90]{fig8c.eps}\\[-7.5mm]
667: %\hspace*{-2mm}\includegraphics[height=5.5cm, angle=-90]{fig8d.eps}}
668: %{%Like Fig.~\ref{fig1a}, but for the payoff matrix
669: %$\vec{P}=(-1,2,-2,1)$ and
670: %$D_1=D_2=0.05$ (top), $D_1=5$, $D_2=0.05$ (second),
671: %$D_1=500$, $D_2=0.05$ (third), $D_1=500$, $D_2=5$ (bottom).}
672: %{fig8a}
673: \par
674: The temporal development of the distribution of entities and of the
675: overall success in the subpopulations gives additional information
676: (see Fig.~\ref{fig9}).
677: As in the case of lane formation, the overall success fluctuates
678: strongly, because the subpopulations do not separate from each other,
679: causing ongoing interactions. Hence, the resulting
680: distribution is not stable, but changes continuously. It can,
681: therefore, happen, that clusters of subpopulation 2 merge, which is
682: associated with an increase of overall success in subpopulation 2
683: (see Fig.~\ref{fig9}).
684: %\par\abb{
685: %\vspace*{-5mm}
686: %\hspace*{6mm}\includegraphics[height=6cm, angle=-90]{fig9a.eps}\\[-5mm]
687: %%\vspace*{-5mm}
688: %\hspace*{-2mm}\includegraphics[height=5.5cm, angle=-90]{fig9b.eps}}
689: %{Top: Temporal evolution of the distribution of entitities within subpopulation
690: %$a=2$ for $\vec{P}=(-1,2,-2,1)$ and $D_1=500$, $D_2=0.05$.}
691: %{fig9a}
692: \par\abb{
693: \vspace*{-5mm}
694: \hspace*{6mm}\includegraphics[height=6cm, angle=-90]{fig9a.ps}\\[-5mm]
695: %\vspace*{-5mm}
696: \hspace*{-2mm}\includegraphics[height=5.5cm, angle=-90]{fig9b.ps}}
697: {Temporal evolution of the distribution of entitities within subpopulation
698: $a=2$ (top) and of the overall successes (bottom)
699: for $\vec{P}=(-1,2,-2,1)$ and $D_1=50$, $D_2=0.5$.}
700: {fig9}
701:
702: \section{\large Conclusions}
703:
704: We have proposed a
705: game theoretical model for self-organization in
706: space, which is applicable to many kinds
707: of biological, economic, and social systems with various types of
708: profitable or competitive self- and cross-interactions
709: between subpopulations of the system. Depending on the structure of
710: the payoff matrix, we found several different self-organization
711: phenomena like agglomeration, segregation, or lane formation. It
712: turned out that medium noise strengths can increase the resulting
713: level of order, while a high noise level leads to more or less homogeneous
714: distributions of entities over the available space. The mechanism of
715: noise-induced ordering in the above discussed systems with
716: short-range interactions seems to be the following: Noise extends a
717: ``pre-ordering'' phase by keeping up a quasi-continuous distribution
718: of entities, which allows a long-range ordering. For asymmetric
719: payoff matrices, we can even have the phenomenon of noise-induced
720: self-organization, although we start with completely homogeneous
721: distributions and homogeneous (translation-invariant) payoffs.
722: However, the phenomenon requires different noise amplitudes in both
723: subpopulations. The role of noise is to suppress agglomeration in one
724: of the subpopulations, in this way reducing repulsive effects that
725: would suppress agglomeration in the other subpopulation.
726: \par
727: We point out that all the above results can be semi-quantitatively understood
728: by means of a linear stability analysis of a related continuous
729: version of the model \cite{helvic}. This continuous version indicates
730: that the linearly most unstable modes are the ones with the shortest
731: wave length, so that\linebreak\newpage
732: \noindent one does not expect a characteristic length scale
733: in the system. This is different from reaction-diffusion systems,
734: where the most unstable mode has a finite wave length, which gives
735: rise to the formation of periodic patterns. Nevertheless, the
736: structures evolving in our model are spatially extended, but
737: non-periodic. The spatial extension is increasing with the fluctuation
738: strength, unless a critical noise amplitude is exceeded.
739: \par
740: For a better agreement with real systems,
741: the model can be generalized in many ways.
742: The entities may perform a biased or unbiased random walk in space.
743: One can allow random jumps to neigboring cells
744: with some prescribed probability.
745: This probability may depend on the subpopulation, and thus we can
746: imitate different mobilities of the considered subpopulations.
747: Evolution is slowed down by introducing a threshold, fixed or random,
748: so that the entities change to other
749: cells only if the differences in the relevant
750: successes are bigger than the imposed threshold.
751: The model can be also generalized to higher dimensions,
752: with expected interesting patterns of self-organized structures.
753: \par
754: In general, the random variables $\xi_\alpha(t)$ in the definition of
755: the success functions can be allowed to have different variances
756: for the considered cell $i$ and the
757: neighboring cells, with the interpretation that the uncertainty
758: in the evaluation of the success in the
759: considered cell is different (e.g. smaller)
760: than that in the neighboring cells.
761: Moreover, the uncertainities can be different for various
762: subpopulations, which could reflect to some extent their different
763: knowledge and %(e.g. social)
764: behavior.
765: \par
766: One can as well study systems with more than two subpopulations,
767: the influence of long-range interactions, etc. The entities
768: can also be allowed to jump to more remote cells.
769: As an example, the following update rule could be implemented:
770: Move entity $\alpha$ from cell $i$ to the cell
771: $(i+l)$ for which
772: \begin{equation}
773: %C(1-N_{i+l}(t)/N_{\rm max}) \cdot d^{|l|} \cdot S_a(i+l,t)
774: S_a^{\prime\prime}(i+l,t) = d^{|l|} c(i+l,t) S_a(i+l,t)
775: \end{equation}
776: is maximal ($|l| = 0,1,...,l_{\rm max})$.
777: If there are $m$ cells in the range $\{(x-l_{\rm
778: max}),\dots,(x+l_{\rm max})\}$
779: with the same maximal value, choose one of them randomly with probability $%
780: 1/m$. According to this, when spontaneously moving to another cell, the
781: entity prefers cells in the neighborhood with higher success. The indirect
782: interaction behind this transition, which is based on the observation or
783: estimation of the success in the neighborhood, is short-ranged if
784: $l_{\rm max} \ll I$, otherwise long-ranged. Herein, $l_{\rm max}$
785: denotes the maximum number of cells which an entity
786: can move within one time step.
787: %The factor $C$ in the above formula is
788: %supposed to avoid that the total number $N_i(t)$ of entities in cell $i$
789: %exceeds a certain maximum number $N_{\rm max}$ (reflecting a maximum density
790: %$\rho_{\rm max}=N_{\rm max}/I$).
791: The factor containing $d$ with $0 < d < 1$ allows to
792: consider that it is less likely to move for large distances, if this is not
793: motivated by a higher success. A value $d<1$ may also reflect the fact that
794: the observation or estimation of the success over large distances becomes
795: more difficult and less reliable.\\[5mm]
796: {\em Acknowledgments:}
797: D.H. thanks E\"ors Szathm\'{a}ry and
798: Tam\'{a}s Vicsek for inspiring discussions and
799: the German Research Foundation (DFG)
800: for financial support by a Heisenberg scholarship.
801: T.P. is grateful to the Alexander-von-Humboldt Foundation
802: for financial support during his stay in Stuttgart.
803:
804: \begin{footnotesize}
805:
806: \begin{thebibliography}{99}
807: %
808: %\bibitem{Weidlich} Weidlich, W. \& Haag, G. {\it Concepts and Models of a
809: %Quantitative Sociology} (Springer, Berlin, 1983), pp. 105ff.
810: %
811: \bibitem{We91} Weidlich, W.: Physics and social science---{T}he approach of
812: synergetics. {\it Physics Reports}, vol. 204, pp. 1-163, 1991
813: %
814: \bibitem{game1} von Neumann, J., Morgenstern, O.: {\it Theory of Games and
815: Economic Behavior.} Princeton University, Princeton, 1944
816: %
817: \bibitem{game2} Axelrod, R., Hamilton, W. D.: The evolution of
818: cooperation. {\it Science}, vol. 211, pp. 1390-1396, 1981
819: %Axelrod, R. {\it The Evolution of Cooperation}
820: %(Basic Books, New York, 1984).
821: %
822: \bibitem{game3} Axelrod, R., Dion, D.: The further evolution of
823: cooperation. {\it Science}, vol. 242, pp. 1385-1390, 1988
824: %
825: \bibitem{gamedyn} Hofbauer, J., Sigmund, K.: {\it The Theory of Evolution
826: and Dynamical Systems.} Cambridge University Press, Cambridge, 1988
827: %
828: \bibitem{Hub} Glance, N. S., Huberman, B. A.: The dynamics of social
829: dilemmas. {\it Scientific American}, vol. 270, pp. 76-81, 1994
830: %
831: \bibitem{quantsoc} Helbing, D.: {\it Quantitative Sociodynamics. Stochastic
832: Methods and Models of Social Interaction Processes.} Kluwer Academics,
833: Dordrecht, 1995
834: %
835: \bibitem{frank} Schweitzer, F. (Ed.): {\it Self-Organization of Complex
836: Structures}. {\it From individual to Collective Dynamics}. Gordon and Breach,
837: Amsterdam, 1997
838: %
839: \bibitem{Lewenstein}
840: Lewenstein, M., Nowak, A., Latan\'{e}, B.: Statistical mechanics of
841: social impact.
842: {\it Physical Review A}, vol. 45, pp. 763-776, 1992
843: %
844: \bibitem{Galam} Galam, S.: Rational group decision making. {\it Physica A},
845: vol. 238, pp. 66-80, 1997
846:
847: \bibitem{book} Helbing, D.: {\it Verkehrsdynamik [Traffic Dynamics].}
848: Springer, Berlin, 1997
849:
850: \bibitem{pre} Helbing, D., Moln\'{a}r, P.: Social force model for
851: pedestrian dynamics. {\it Physical Review E}, vol. 51, pp. 4282-4286, 1995
852:
853: \bibitem{phasediag} Helbing, D., Hennecke, A., Treiber, M.:
854: Phase diagram of traffic states in the presence of inhomogeneities.
855: {\it Physical Review Letters}, vol. 82, pp. 4360-4363, 1999
856:
857: \bibitem{solid} Helbing, D., Huberman, B. A.: Coherent moving states in
858: highway traffic. {\it Nature}, vol. 396, pp. 738-740, 1998
859:
860: \bibitem{scatter} Treiber, M. and Helbing, D.:
861: Macroscopic simulation of widely scattered synchronized traffic
862: states. {\it Journal of Physics A: Mathematical and General},
863: vol. 32, pp. L17-L23, 1999
864:
865: \bibitem{opus} Treiber, M., Hennecke, A., Helbing, D.:
866: Congested traffic states in empirical observations and microscopic simulations.
867: Preprint {\tt http://
868: xxx.lanl.gov/abs/cond-mat/0002177} %,
869: submitted to
870: {\it Physical Review E}, 2000.
871:
872: \bibitem{trail} Helbing, D., Keltsch, J., Moln\'{a}r, P.: Modelling the
873: evolution of human trail systems. {\it Nature}, vol. 388, pp. 47-50,
874: 1997
875:
876: \bibitem{Haken} Haken, H. {\it Synergetics.} Springer, Berlin, 1977
877:
878: \bibitem{Adv} Haken, H. {\it Advanced Synergetics.}
879: Springer, Berlin, 1983
880:
881: \bibitem{seg3} Helbing, D., Mukamel, D., Sch\"utz, G. M.: Global phase
882: diagram of a one-dimensional driven lattice gas. {\it Physical Review
883: Letters}, vol. 82, 10-13, 1999
884:
885: \bibitem{manneville}
886: Manneville, P.:
887: {\it Dissipative Structures and Weak Turbulence.}
888: Academic Press, New York, 1990
889: %
890: %\bibitem{Collet} J.-P. Eckmann {\it et al.},
891: %%, H. Epstein, and C. E. Wayne,
892: %Annales de l'Institut Henri Poincare Physique Theorique {\bf 58},
893: %287 (1993);
894: %%\bibitem{Kura}
895: %Y. Kuramoto, Progress in Theor. Phys. Suppl. {\bf 99},
896: %244 (1989).
897:
898: \bibitem{Zee77}
899: Zeeman, E. C. (Ed.): {\it Catastrophe Theory}.
900: Addison-Wesley, London, 1977
901:
902: \bibitem{Ber68}
903: von Bertalanffy, L.: {\it General System Theory}.
904: Braziller, New York, 1968
905:
906: \bibitem{Bu67}
907: Buckley, W.: {\it Sociology and Modern Systems Theory}.
908: Prentice-Hall, Englewood Cliffs, NJ, 1967
909:
910: \bibitem{Rap86}
911: Rapoport, A.: {\it General System Theory. Essential Concepts and
912: Applications}. Abacus Press, Tunbridge Wells, Kent, 1986
913:
914: \bibitem{Ebeling} Feistel, R., Ebeling, W.: {\it Evolution of Complex
915: Systems.} Kluwer Academic, Dordrecht, 1989
916:
917: \bibitem{gd} Helbing, D.: Stochastic and Boltzmann-like models for
918: behavioral changes, and their relation to game theory. {\it Physica A},
919: vol. 193, pp. 241-258, 1993
920:
921: \bibitem{helvic} Helbing, D., Vicsek, T.: Optimal self-organization.
922: {\it New Journal of Physics}, vol. 1, pp. 13.1--13.17, 1999
923:
924: \bibitem{stochres} Gammaitoni, L., H\"anggi, P., Jung, P., Marchesoni,
925: F.: Stochastic resonance. {\it Review of Modern Physics}, vol. 70,
926: pp. 223-288, 1998
927:
928: \bibitem{biol1a} {\L}uczka, J., Bartussek, R., H\"anggi, P.: White noise
929: induced transport in periodic structures. {\it Europhysics Letters}, vol. 31,
930: pp. 431-436, 1995
931: %
932: %\bibitem{biol1b} Bartussek, R., Reimann, P. \& H\"anggi, P. Precise
933: %numerics versus theory for correlation ratchets. {\it Phys. Rev. Lett.} {\bf %
934: %76}, 1166--1169 (1996).
935:
936: \bibitem{biol1c} Reimann, P., Bartussek, R., H\"au{\ss}ler, R., H\"anggi,
937: P.: Brownian motors driven by temperature oscillations. {\it Physics Letters A},
938: vol. 215, pp. 26-31, 1996
939:
940: \bibitem{freezing} Helbing, D., Farkas, I. J., Vicsek, T.:
941: Freezing by heating in a driven mesoscopic system.
942: {\it Physical Review Letters}, vol. 84, pp. 1240-1243, 2000
943:
944: \bibitem{Pri} Nicolis, G., Prigogine, I.: {\it Self-Organization in
945: Nonequilibrium Systems. From Dissipative Structures to Order through
946: Fluctuations.} Wiley, New York, 1977
947:
948: \bibitem{Pri2} Prigogine, I.: Order through fluctuation:
949: Self-organization and social system.
950: Jantsch, E. and Waddington, C. H. (Eds.): {\it Evolution and
951: Consciousness. Human Systems in Transition}, pp. 93-130.
952: Addison-Wesley, Reading, MA, 1976
953:
954: \bibitem{HoLe84}
955: Horsthemke, W., Lefever, R.: {\it Noise-Induced Transitions}.
956: Springer, Berlin, 1984
957: %
958: %\bibitem{RiVo89}
959: %H. Risken and H.~D. Vollmer (1989) Methods for solving {F}okker-{P}lanck
960: % equations with applications to bistable and periodic potentials.
961: %In: F. Moss and P.~V.~E. McClintock, eds. {\it Noise in Nonlinear Dynamical
962: % Systems, Vol.~1: {T}heory of Continuous {F}okker-{P}lanck Systems}.
963: %Cambridge University Press, Cambridge.
964:
965: \bibitem{turing} Turing, A. M.: The chemical basis of morphogenesis.
966: {\it Philosophical Transactions of the Royal Society of London},
967: vol. B237, pp. 37-72, 1952
968:
969: \bibitem{turing1}
970: Murray, J. D.: {\it Lectures on Nonlinear Differential Equation-Models in
971: Biology.} Clanderon Press, Oxford, 1977
972: %Murray, J. D. {\it Mathematical Biology.} Springer, Berlin, 1989
973:
974: %\bibitem{turing1} Othmer, H. G., Current problems in pattern formation,
975: %in {\it Lectures on Mathematics in the Life Sciences}, Vol. 9,
976: %edited by S. Levin, AMS, 1977
977:
978: \bibitem{turing2}
979: Fife, P. C.: {\it Mathematical aspects of reacting and diffusing
980: systems.}
981: % Lect. Notes in Biomathematics,
982: Springer, New York, 1979
983:
984: \bibitem{turing3}
985: Convay, E., Hoff, D., Smoller, J. A.: Large time behavior of systems of
986: nonlinear diffusion equations.
987: {\it SIAM Journal of Applied Mathematics}, vol. 35, pp. 1-16, 1978
988: %
989: %\bibitem{turing4}
990: %Levin, S. A., Population models and community structure in heterogeneous
991: %enviroments, Math. Assoc. of America Study, in Math. Biol. Vol. II,
992: %Populations and Communities, S. Levin ed. 1977
993: %
994: %\bibitem{turing5}
995: %Evans, G. T., Diffusive structure: counterexample to any explanation? J.
996: %Theor. Biol. 92 (1980) 313-315
997:
998: \bibitem{turing4} Kessler, D. A., Levine, H.: Fluctuation-induced
999: diffusive instabilities. {\it Nature}, vol. 394, pp. 556-558, 1998
1000:
1001: \bibitem{turing5} Zhonghuai, H., Lingfa, Y., Zuo, X., Houwen, X.:
1002: Noise induced pattern transition and spatiotemporal
1003: stochastic resonance.
1004: {\it Physical Review Letters}, vol. 81, pp. 2854-2857, 1998
1005:
1006: \bibitem{gran1} Rosato, A., Strandburg, K. J., Prinz, F., Swendsen,
1007: R. H.:
1008: Why the Brazil nuts are on top: Size segregation of particulate matter by
1009: shaking.
1010: {\it Physical Review Letters}, vol. 58, pp. 1038-1041, 1987
1011:
1012: \bibitem{gran2} Gallas, J. A. C., Herrmann, H. J., Soko{\l}owski, S.:
1013: Convection cells in vibrating granular media.
1014: {\it Physical Review Letters}, vol. 69, pp. 1371-1374, 1992
1015:
1016: \bibitem{gran3} Umbanhowar, P. B., Melo, F., Swinney, H. L.:
1017: Localized excitations in a vertically vibrated granular layer.
1018: {\it Nature}, vol. 382, pp. 793-796, 1996
1019: %
1020: %\bibitem{Schw91}
1021: %Schweitzer, F., Bartels, J., Pohlmann, L. Simulation of opinion
1022: % structures in social systems.
1023: %In: W. Ebeling, M. Peschel, and W. Weidlich (Eds.) {\it Models of
1024: % Selforganization in Complex Systems (MOSES)}.
1025: %Akademie Verlag, Berlin, 1991
1026:
1027: \bibitem{Keizer} Keizer, J.: {\it Statistical Thermodynamics of
1028: Nonequilibrium Processes.} Springer, New York, 1987
1029:
1030: \bibitem{Hel93} Helbing, D.: Boltz\-mann-like and Boltz\-mann-Fokker-Planck
1031: equations as a foundation of behavioral models. {\it Physics A}, vol. 196,
1032: pp. 546-573, 1993
1033:
1034: \bibitem{granular} Santra, S. B., Schwarzer, S., Herrmann, H.:
1035: Fluid-induced particle-size segregation in sheared granular
1036: assemblies.
1037: {\it Physical Review E}, vol. 54, 5066-5072, 1996
1038:
1039: \bibitem{eshel} Ben-Jacob, E.,
1040: Schochet, O., Tenenbaum, A., Cohen, I., Czir\'{o}k, A., Vicsek, T.:
1041: Generic modelling of cooperative growth patterns in bacterial
1042: colonies. {\it Nature}, vol. 368, pp. 46-49, 1994
1043:
1044: \bibitem{jacob2} Ben-Jacob, E.: From snowflake formation to growth of
1045: bacterial colonies, Part II: Cooperative formation of complex colonial
1046: patterns. {\it Contemporary Physics}, vol. 38, pp. 205-241, 1997
1047:
1048: \bibitem{biol6} Kessler, D. A., Levine, H.: Pattern for-\linebreak\newpage
1049: \noindent mation in {\it
1050: Dictyostelium} via the dynamics of cooperative biological entities. {\it
1051: Physical Review E}, vol. 48, 4801-4804, 1993
1052:
1053: \bibitem{biol2} Vicsek, T.,
1054: Czir\'ok, A., Ben-Jacob, E., Cohen, I., Shochet, O.:
1055: Novel type of phase transition in a system of self-driven particles.
1056: {\it Physical Review Letters}, vol. 75, pp. 1226-1229, 1995
1057:
1058: \bibitem{biosys} Schweitzer, F., Lao, K., Family, F.: Active random walkers
1059: simulate trunk trail formation by ants. {\it BioSystems}, vol. 41, 153-166,
1060: 1997
1061:
1062: \bibitem{millonas} Rauch, E. M., Millonas, M. M., Chialvo, D. R.:
1063: Pattern formation and functionality in swarm models. {\it Physics
1064: Letters A}, vol. 207, 185-193, 1995
1065:
1066: \bibitem{Wolfram} Wolfram, S.: Cellular automata as models of complexity.
1067: {\it Nature}, vol. 311, pp. 419-424, 1984
1068:
1069: \bibitem{Stauffer} Stauffer, D.: Computer simulations of cellular automata.
1070: {\it Journal of Physics A: Mathematical and General}, vol. 24, pp. 909-927, 1991
1071:
1072: \bibitem{Bernardo} Huberman, B. A., Glance, N. S.: Evolutionary games and
1073: computer simulations. {\it Proceedings of the National Academy of
1074: Science USA}, vol. 90, pp. 7716-7718, 1993
1075:
1076: \bibitem{Sche71}
1077: Schelling, T.~C.: Dynamic models of segregation.
1078: {\it Journal of Mathematical Sociology}, vol. 1, pp. 143-186, 1971
1079: \end{thebibliography}
1080: \end{footnotesize}
1081: \end{document}
1082:
1083:
1084:
1085: \bibitem{Info} Haken, H. {\it Information and Self-Organization} (Springer,
1086: Berlin, 1988).
1087:
1088: \bibitem{evol} Eigen, M. \& Schuster, P. {\it The Hypercycle} (Springer,
1089: Berlin, 1979).
1090:
1091: \bibitem{biol3} Albano, E. V. Self-organized collective displacements of
1092: self-driven individuals. {\it Phys. Rev. Lett.} {\bf 77}, 2129--2132 (1996).
1093:
1094: \bibitem{biol4} Bussemaker, H. J., Deutsch, A. \& Geigant, E. Mean-field
1095: analysis of a dynamical phase transition in a cellular automaton model for
1096: collective motion. {\it Phys. Rev. Lett.} {\bf 78}, 5018--5021 (1997).
1097:
1098: \bibitem{biol5} Toner, J. \& Tu, Y. Long-range order in a two-dimensional
1099: dynamical {\it XY} model: How birds fly together. {\it Phys. Rev. Lett.}
1100: {\bf 75}, 4326--4329 (1995).
1101:
1102: \bibitem{biol6} Kessler, D. A. \& Levine, H. Pattern formation in {\it \
1103: Dictyostelium} via the dynamics of cooperative biological entities. {\it %
1104: Phys. Rev. E} {\bf 48}, 4801--4804 (1993).
1105:
1106: \bibitem{Drasdo} Drasdo, D., Kree, R. \& McCaskill, J. S., A Monte Carlo
1107: approach to tissue-cell populations. {\it Phys. Rev. E} {\bf 52}, 6635--6657
1108: (1995).
1109:
1110: \bibitem{ants1} Schweitzer, F., Lao, K. \& Family, F. Active random walkers
1111: simulate trunk trail formation by ants. {\it BioSystems} {\bf 41}, 153--166
1112: (1997).
1113:
1114: \bibitem{ants2} Schweitzer, F., Ebeling, W. \& Tilch, B. Complex motion of
1115: Brownian particles with energy depots. {\it Phys. Rev. Lett.} {\bf 80},
1116: 5044--5047 (1998).
1117:
1118: %\bibitem{mandel} Mandelbrot, B. The variation of certain speculative
1119: %prices. {\it Journal of Business} {\bf 35}, 394--419 (1963).
1120:
1121: %\bibitem{Bouchaud} Laloux, L., {\it et al.} Are financial crashes
1122: %predictable? {\it Europhys. Lett.} {\bf 45}, 1--5 (1999).
1123:
1124: %Bouchaud, J.-P. \& Sornette, D. The Black-Scholes option
1125: %pricing problem in mathematical finance: Generalization and extensions for a
1126: % large class of stochastic processes.
1127: %{\it J. Phys. I France}, {\bf 4} 863--881 (1994).
1128:
1129: %\bibitem{solomon} Levy, M., Levy, H. \& Solomon, S. Microscopic simulation
1130: %of the stock market: The effect of microscopic diversity. {\it J. Phys. I
1131: %(France)} {\bf 5}, 1087--1107 (1995).
1132:
1133: %\bibitem{stanley} Mantegna, R. N. \& Stanley, E. Scaling behavior in the
1134: %dynamics of an economic index. {\it Nature} {\bf 376}, 46--49 (1995).
1135:
1136: %\bibitem{peinke} Ghashgaie, S., {\it et al.} Turbulent cascades in foreign
1137: %exchange markets. {\it Nature} {\bf 381}, 767--770 (1996).
1138:
1139: %\bibitem{zhang} Caldarelli, G., Marsili, M. \& Zhang, Y.-C. A prototype
1140: %model of stock exchange. {\it Europhysics Letters} {\bf 40}, 479--484 (1997).
1141:
1142: %\bibitem{Stauf} de Oliveira, S. M., de Oliveira, M. C. \& Stauffer, D. {\it %
1143: %Evolution, Money, War, and Computers---Non-Traditional Applications of
1144: %Computational Statistical Physics} (Teubner, Stuttgart, 1999).
1145:
1146: \bibitem{jacob1} %Ben-Jacob, E. \& Garik, P.
1147: %The information of patterns in non-equilibrium growth.
1148: %{\it Nature} {\bf 343}, 523--530 (1990).
1149: Ben-Jacob, E. From snowflake formation to the growth of bacterial colonies,
1150: Part I: Diffusive patterning in non-living systems. {\it Contemporary Physics%
1151: } {\bf 34}, 247--273 (1993).
1152:
1153: %\bibitem{game4} Maynard Smith, J. \& Price, G. The logic of animal
1154: %conflicts. {\it Nature} {\bf 246}, 15--18 (1973).
1155:
1156: \bibitem{Nowak} Nowak, M. A. \& Sigmund, K. Tit for tat in heterogeneous
1157: populations. {\it Nature} {\bf 355}, 250--253 (1992).
1158:
1159: \bibitem{life} Bak, P., Chen, K. \& Creutz, M. Self-organized criticality
1160: in the 'Game of Life'. {\it Nature} {\bf 342}, 780--782 (1989).
1161:
1162: \bibitem{may} Nowak, M. A. \& May, R. M. Evolutionary games and spatial
1163: chaos. {\it Nature} {\bf 359}, 826--829 (1992).
1164:
1165: \bibitem{freezing} Helbing, D., Farkas, I. \& Vicsek, T. Freezing by
1166: heating in a driven mesoscopic system. Preprint
1167: http://xxx.lanl.gov/abs/cond-mat/9904326.
1168:
1169: \bibitem{complex} Helbing, D. A fluid-dynamic model for the movement of
1170: pedestrians. {\it Complex Systems} {\bf 6}, 391--415 (1992).
1171:
1172: %\bibitem{Rayleigh} Lord Rayleigh, {\it Proc. Math. Soc. London} {\bf 4},
1173: %357ff (1873).
1174:
1175: %\bibitem{Onsager} Onsager, L. Reciprocal relations in irreversible
1176: %processes I, {\it Phys. Rev.} {\bf 37}, 405--426 (1931).
1177:
1178: %\bibitem{deGroot} de Groot S. R. \& Mazur, P. {\it Non-Equilibrium
1179: %Thermodynamics} (North-Holland, Amsterdam, 1962).
1180:
1181: \bibitem{Graham} Graham, R. \& T\'{e}l, T. Macroscopic potentials of
1182: dissipative dynamical systems. Albeverio, S., Blanchard, P., Streit,
1183: L. (Eds.) {\it Stochastic Processes and Their
1184: Applications} (Kluwer,
1185: Dordrecht, 1990), pp.~153--175. %
1186:
1187: %\bibitem{Callen} Callen, H. B. {\it Thermodynamics} (Wiley, New York,
1188: %1960).
1189:
1190: \bibitem{active} F. Schweitzer and L. Schimansky-Geier,
1191: Clustering of ``active'' walkers in a two-component system.
1192: {\it Physica A} {\bf 206}, 359--379 (1994).
1193:
1194: \bibitem{SOC} Bak, P., Tang, C. \& Wiesenfeld, K. Self-organized
1195: criticality: An explanation of $1/f$ noise. {\it Phys. Rev. Lett.} {\bf 59},
1196: 381--384 (1987).
1197:
1198: \bibitem{seg1} Schmittmann, B. \& Zia, R. K. P. {\it Statistical Mechanics
1199: of Driven Diffusive Systems}, Vol.~17 of {\it Phase Transitions and Critical
1200: Phenomena}, edited by C. Domb \& J. L. Lebowitz (Academic Press, London,
1201: 1995).
1202:
1203: \bibitem{seg2} Evans, M. R., Kafri, Y., Koduvely, H. M. \& Mukamel, D.
1204: Phase separation in one-dimensional driven diffusive systems. {\it Phys.
1205: Rev. Lett.} {\bf 80}, 425--429 (1998).
1206:
1207: %\bibitem{Weidmann} Weidmann, U. {\it {T}ransporttechnik der {F}u{\ss}g{\"a}%
1208: %nger}. Vol.~90 of the {\it Schriftenreihe des IVT (Institut f{\"u}r
1209: %Verkehrsplanung, Transporttechnik, Stra{\ss}en- und Eisenbahnbau)}. (ETH Z{%
1210: %\"u}rich, ETH-H{\"o}nggerberg, CH-8093 Z{\"u}rich, 1993).
1211:
1212:
1213:
1214:
1215: \bibitem{Weidlich1} Weidlich, W. {\it to appear } (Springer, Berlin, 2000)
1216:
1217: \bibitem{helpla} Helbing D., Platkowski T., work in progress.
1218:
1219: \end{thebibliography}
1220:
1221: %\end{small}
1222: %\end{scriptsize}
1223: \end{footnotesize}
1224:
1225: %\end{bibliography}
1226: \clearpage
1227:
1228:
1229:
1230: \end{document}
1231: \par
1232: Overcrowding effects can be easily included into the dynamics
1233: defined in Section 2. For attractive interactions, for example,
1234: this can be done by defining the left and right neighbors'
1235: successes by %$S_{\mp}(i,t)$
1236: \begin{equation}
1237: S_{\mp}(i,t)=c_{\mp}[S_a(i\mp 1,t)+d \cdot S_a(i\mp 2,t)] \, ,
1238: \end{equation}
1239: where the functions $c_{-}$ and $c_{+}$ describe a saturation effect:
1240: \begin{equation}
1241: c_{\mp}=1-\frac{n_{i\mp 1}(t)} {n_{\rm max}}, \ \
1242: %\begin{equation}
1243: n_.(t)=n^1_.(t)+n^2_.(t).
1244: %\end{equation}
1245: \end{equation}
1246: $c_{\mp}$ are supposed to prevent that the total number of entities
1247: in cell $i$ becomes bigger then a
1248: certain prescribed maximum number
1249: $n_{\rm max}=\rho_{\rm max}I$, where $\rho_{\rm max}$ is the maximum density.
1250: \par
1251: In the figures discussed below, the left diagram of each composed
1252: figure is referred to as ``a'',
1253: the right one as ``b''.
1254: In the diagrams illustrating the final evolving state, the cells are
1255: represented by the numbers on the $x$-axis, while the $y$-axis (or $z$-axis)
1256: displays the number of entities in the cells. The sticks and bars
1257: correspond to the subpopulations $a=1$ and $a=2$, respectively.
1258: Their heights represent the corresponding occupation numbers.
1259: If not stated otherwise, the figures discussed below illustrate the behavior
1260: of systems without overcrowding effects and without interactions with
1261: next-nearest neighbors (i.e. $d=0$).
1262: $D_a$ represents the %maximum
1263: amplitude of the fluctuations of the success functions.
1264: