Communications 
Cécile APPERTROLLAND

The response of a highway to a local perturbation
(onramp) is taken as a test for lane changing rules.
In a first stage, real data were analyzed, with
a special focus on the relaxation length, i.e. the
length necessary to obtain a new equilibrium among
the lanes, downwards the onramp.
In a second stage, the setup is reproduced with
numerical simulations based on cellular automata.
The analysis of the macroscopic behavior of
the system downwards the onramp allows to point
some limitations of the simulation approach.

Macroscopic relaxation after onramps:
real data vs. cellular automata simulations

Marc BARTHELEMY


Empirical properties of transportation networks

Laure BOURGOIS

Accident analysis highlights that most accidents involving
pedestrians take place at crosswalk, shared space where potential
conflicts between pedestrians and drivers can occur. The context of
these works aims to contribute to the pedestrian modelling in
crosswalk situations and more generally in urban areas.
Thus we developed SPUREN (Simulation of Pedestrian in Urban
Environment) a software prototype for the simulation, at microscopic
level, of both pedestrians and vehicles in an urban context.
For this purpose, we have used the Social Force Model (SFM), model
developed by Helbing, and widely used for crowd simulation.
However,the limit of this model is to consider pedestrians only in
homogeneous situations, i.e. pedestrians interacting with both
pedestrians and fixed obstacles (wall, building...). Thus
interactions with vehicles are not considered. So, we have extended
this model by incorporating both a psychological decision process of
crossing the street based on perceptual features, and a mechanism of
anticipation based on the estimated crossing time. We used an
agentbased approach, by describing pedestrians and vehicles as a set
of properties and perceptive and anticipative behaviours.

Pedestrians' crossing model based on perceived environment and
psychological process

Christine BUISSON

The communication will summarize some of the recent findings
obtained after detailed observation of trajectories data. A
sort bibliographic part will recall that heterogeneities among
drivers ?vehicles units are suspected to be at the origin of
some of the main disturbances in traffic flow, like
oscillations, capacity drop? We will then show how trajectories
analysis permits to improve the understanding of this
phenomenon. We will then show the limits of the present result
and propose an experimental framework to complete the data
sets.

Impact of heterogeneities on the global behavior of road traffic

Hugues CHATE



Mohcine CHRAIBI

The first part gives a brief overview of different experiments and
measurement methods providing results on an individual scale. In the
second part a spatially continuous forcebased model for simulating
pedestrian dynamics is introduced which includes an elliptical volume
exclusion of pedestrians.

Experimental validation of forcebased models for pedestrian dynamics

Winnie DAAMEN

The ability of microscopic (simulation) models to represent
lane changing behavior according to reality has recently been
questioned. In this paper the merging maneuver (a specific
type of lane changing) is analyzed using empirical data. First,
a conceptual model has been composed, including the factors
influencing merging behavior, namely the merge location and its
relation to prevailing driving conditions, gap acceptance and
the relaxation phenomenon. The empirical dataset consists of 35
minutes covering 400 meters of freeway, collected using a
camera mounted underneath a helicopter. This results in a
dataset of 3459 vehicles trajectories, from which 704
trajectories describe merging vehicles. It is found that
different merge locations are used under congested and
freeflow traffic conditions. During freeflow, most vehicles
merge at the first half of the acceleration lane. Under
congested traffic conditions, relatively more merges are
registered at the end of the acceleration lane. The smallest
accepted gap observed in the dataset lies between 0.75 s and
1.0 s. Net headways between the merging vehicle and his new
leader and new follower of less than 0.25 s are recorded. These
very short accepted gaps are growing over time, indicating
relaxation behavior.
From the data analysis it can be concluded that gap acceptance
theories, as they are used in current models and theories to
model merge behavior, are not able to model the observed
behavior accurately.

Empirical analysis of merging behaviour at freeway onramps and
consequences for modelling

Olivier DAUCHOT

In many interesting situations, the interactions among selfpropelled
agents lead to the spontaneous emergence of selforganized collective
motion. The ubiquity of the phenomenon at all scales raises the
question of the existence of some underlying universal mechanisms.
Recent numerical and analytical studies have confirmed the existence
of a transition from a disordered state at large noise to a state
with various collective properties reflecting the local symmetry of
the particles and their interactions. Though, there are still very
few experimental situations where the onset of collective motion can
be attributed to spontaneous symmetry breaking.
Here, we report on experiments conducted with both polar self
propelled and apolar Brownian disks and by comparing the dynamics of
both systems in the same experimental conditions, we demonstrate
without ambiguity that collective motion emerges from the interplay
of selfpropulsion and hardcore repulsion only [1]. Interestingly
the alignment, which has no nematic origin, is effectively induced
during the collisions because of the self propulsion.
[1] Phys. Rev. Lett 105 135702 (2010)

Collective motion of vibrated polar granular disks.

David GUILBERT

The communication presents the experimental platform SAROT (Site Angevin de Référence
pour l’Observation du Trafic) which is composed of various sensors such as loops,
lasers, cameras. From this platform, databases are collected for traffic analysis.
Different applications experimented on this platform will be discuss.

Presentation of an experimental platform and applications

Henk HILHORST

We consider a twolane road with two opposite traffic directions and
the possibility of platoon formation and vehicles overtaking one
another.
We model this road on the basis of a minimal set of assumptions.
Within the context of this model, as the traffic intensity on both
lanes increases, a symmetry breaking occurs: there is a slow lane
where long platoons form behind the slowest vehicles, and a fast lane
where overtaking is easy due to the wide spacing between the slow
platoons in the opposite direction.

A minimal model for twolane bidirectional overtaking traffic

Serge HOOGENDOORN

This paper proposes a new datadriven stochastic carfollowing model
based on the principles of psychospacing or actionpoint modeling.
It uses empirical or experimental trajectory data and mimics the main
microscopic behavioral characteristics present in the data.
In the actionpoint model, regions are defined in the {\em relative
speed  distance headway} plane in which the follower is likely to
perform an action (increase or decrease acceleration) or not. These
regions can be established empirically from vehicle trajectory data,
yielding a joint cumulative probability distribution function of the
action points. Furthermore, the conditional distribution of the
actions (the size of the acceleration or deceleration given the
current distance headway and relative speed or given the acceleration
before the action) can be determined from these data as well.
To assess the data correctly, a new filtering technique is proposed.
The main hypothesis behind this idea is that the speed profile is a
continuous piecewise linear function: accelerations are piecewise
constant changing values at nonequidistant discrete time instants.
The durations of these constant acceleration periods are not fixed,
but depend on the state of the follower in relation to its leader.
The data analysis indeed illustrates that driving behavior shows
nonequidistant constant acceleration periods.
The aforementioned distributions of the action points and the
conditional accelerations form the core of the presented datadriven
stochastic model. The paper depicts the mathematical formalization
describing how these distributions can be used to simulate
carfollowing behavior. Based on empirical data collected on a Dutch
motorway, we illustrate the workings of the approach and the
simulation results.

Wiedemann revisited: A New Trajectory Filtering Technique and its
Implications for CarFollowing Modeling

Asja JELIC

Human crowds and pedestrian groups exhibit complex and
coordinated spatiotemporal patterns such as the spontaneous
spatial organization of pedestrian flows into lines, and the
oscillations of fluxes at gates or intersections. Despite their
importance, these phenomena are not well understood, in
particular the `microscopic' interactions between the
individuals and with their environment which govern the
macroscopic behavior at medium and high densities.
In the frame of a collective project implying four French
laboratories (LPT in Orsay, CRCA and IMT in Toulouse, BUNRAKU
in Rennes), we have started an experimental and theoretical
study of the formation of spatiotemporal structures within
moving pedestrians crowds. We shall present the first results
from the experimental campaign of 2009. Our aim is to better
understand the role of the various (physical and behavioral)
parameters which control and modulate these structures in
controlled laboratory conditions, and to develop realistic
analytical and simulation models of crowds based on these
experimental data.

Dynamics of pedestrians: crowds and individuals

JeanPatrick LEBACQUE

The presentation addresses first static assignment, with emphasis on
the principles of user choice and utility. The main problem
formulations: fixed point, variational inequalities and dynamic
system, will be described. The specific difficulties related to
multimodality will be stated. Second the communication addresses
predictive dynamic assignment, focusing on dynamic system and field
approaches. An example based on multiagents and the crossentropy
approch will be described.
Authors for the communication:
J.P. Lebacque(1), M.M. Khoshyaran (2), T.Y. Ma(3)
(1) GRETIA  INRETS
(2) ETC Economics Traffic Clinic
(3) LET  ISH

A short survey of assignment in transportation networks

JeanPierre NADAL


Reaction times in simple decision tasks: from neural modelling to
psychophysics

Bertrand MAURY

We propose a general framework to incorporate congestion in the
modeling of crowd motion in evacuation situations. In its simpler,
microscopic form, the approach we propose is based on the definition
of a desired velocity (corresponding to the velocity one would have
in the absence of others); the actual velocity is then defined as the
projection of this desired velocity onto the set of feasible
velocities (velocity which do not violate the nonoverlapping
constraints between individuals). We proposed recently a macroscopic
version of this approach, which raises several issues from the
theoretical, numerical, and modeling standpoints. In particular, the
difference between both models sheds light on the importance of the
microscopic arrangements between individuals upstream an exit door
in highly packed situations.

Handling of congestion in crowd motion modeling

Pascal PANIZZA

Understanding the flow of discrete elements through networks is of
importance for diverse phenomena, including microfluidics for
controlled droplet traffic, blood flows for functioning
cardiovascular systems, and even for road traffic for optimized road
networks. Addressing this issue requires a description of the
mechanisms that govern flow partitioning at a node. In most case,
splitting rules arte complex since they often involve either human
decision making or noise. droplet traffic thus appears as a model
system since a droplet reaching a node simply flows through the arm
having the largest volumetric flow rate. Despite this robust and
simple rule this paradigmatic system exhibits complex dynamics:
numerous bifurcations between periodic regimes as well as
multistability are observed. Our aim is to present an overview of
recent results obtained on droplet traffic in microfluidic devices.

Droplet traffic in microfluidic networks

Fernando PERUANI

Selfpropelled particle systems are ubiquitous in nature, flocks of
birds, school of fish, and human crows are just a few of the many
existing examples. One relevant question is how information spreads
in these highly dynamical adhoc networks. As one can presume, the
information propagation depends on the motility pattern exhibited by
the agents, which can be highly ordered as in a flock of birds, or
poorly organized as in a crowd.
We will start with the simple case where particles perform a simple
Brownian motion, and learn how the speed of the particles affects the
way information spreads [1,2,3]. We will then focus on a coherently
moving flock to find that statistical properties of the information
spreading are significantly different from the classical contact
process [3]. We will see that the ordering of particles, i.e., the
emergence of common moving direction, facilitates the spreading of
information.
[1] Phys. Rev. Lett. 100, 168103 (2008).
[2] IEEE JSAC 28, 524531 (2010)
[3] Peruani and Chate in preparation.

Information spreading in selfpropelled particle systems

Julien PETTRE

In the everyday exercise of controlling their locomotion, humans rely
on their optic flow of the perceived environment to achieve
collisionfree navigation. In crowds, in spite of the complexity of
the environment made of numerous obstacles, humans demonstrate
remarkable capacities in avoiding collisions. Cognitive science work
on human locomotion states that relatively succinct information is
extracted from the optic flow to achieve safe locomotion. In this
paper, we explore a novel visionbased approach of collision
avoidance between walkers that fits the requirements of interactive
crowd simulation. By simulating humans based on cognitive science
results, we detect future collisions as well as the level of danger
from visual stimuli. The motorresponse is twofold: a reorientation
strategy prevents future collision, whereas a deceleration strategy
prevents imminent collisions. Several examples of our simulation
results show that the emergence of selforganized patterns of walkers
is reinforced using our approach. The emergent phenomena are visually
appealing. More importantly, they improve the overall efficiency of
the walkers? traffic and avoid improbable locking situations.

A SyntheticVisionBased Steering Approach for Crowd Simulation

Ludger SANTEN

We introduce a model for active transport on inhomogeneous networks
embedded in a diffusive environment which is motivated by vesicular
transport on actin filaments. In the presence of a hardcore
interaction, particle clusters are observed that exhibit an
algebraically decaying distribution in a large parameter regime,
indicating the existence of clusters on all scales. The scalefree
behavior can be understood by a mechanism promoting preferential
attachment of particles to large clusters. The results are compared
with a diffusionlimited aggregation model and active transport on a
regular network. For both models we observe aggregation of particles
to clusters which are characterized by a finite size scale if the
relevant time scales and particle densities are considered.

Active transport and cluster formation on disordered networks

Andreas SCHADSCHNEIDER

We report results from largescale experiments of pedestrian
dynamics. Different scenarios are investigated, e.g. fundamental
diagrams in corridors, bottlenecks and motion around corners. We also
discuss the implications of these experiments for the modelling of
pedestrian dynamics, especially for cellular automata approaches.

Pedestrian dynamics: From experiments to models

Martin TREIBER

In spite of investigating traffic flow dynamics for decades,
some fundamental questions are not yet settled: Can the
multitude of observed spatiotemporal patterns of congested
traffic be decomposed into elementary patterns? If so, into
how many patterns, and what are their defining properties? Are
there two or three traffic phases, or is the concept of a
``traffic phase'' meaningless? Of which type are instabilities
of traffic flow? In this contribution we want to elucidate
these questions with the combined force of empirical data,
models, simulations, and analytic calculations. Analyzing
detector data from several hundred traffic jams on freeways in
several countries, we found that nearly all of them are
consistent with linear instability of the upstreamconvective
type, i.e., isolated perturbations grow but eventually leave
the investigated area in the upstream direction. Based on this
type of instability and sustained localized perturbations
(e.g., lane changings near ramp regions), we develop a generic
dynamic model in terms of stochastic differential equations
whose parameters are analytically related to that of
microscopic and macroscopic models. The approximate analytic
solutions of the generic model are nearly identical to
simulations of the respective microscopic or macroscopic models
and consistent with observations. This gives evidence that
traffic instabilities may be a product of convective
instabilities and sustained perturbations.

Tackling traffic flow dynamics with the combined force of
empirical data, models, simulations, and analytics

Véronique VEQUE

A Vehicular AdHoc Network, or VANET, is a technology that uses
moving cars as nodes in a network to create a mobile ad hoc network.
In a VANET, each car is supposed to be equipped with a wireless
network interface, and each car acts as a wireless router. In this
presentation, we present VANET, its most promising applications, and
related research domains which are still open: unicast routing,
dissemination of message, security, etc. We focus on two specific
problems and propose two associated protocols. The first problem
takes place in Internetbased applications between a source and a
destination where a client in the vehicle needs to access a server in
Internet. This type of communication needs first to communicate with
the Internet gateway. In the VANET, we have thus to provide a unicast
routing scheme to find a path to reach a destination through
intermediate vehicles used as routers/forwarders. The second one
intends to improve safety on road by broadcasting alert messages to
all the vehicles on a given highway/road section. We show that
requirements for these both applications are very different. We
discuss the impact of vehicles traffic (vehicle density and
distributions) on these protocol performances.

Impact of vehicular traffic on some VANET applications

Jonathan WARD

Motivated by inductance loop data, I will speak about some
recent results in the classification of the (linear) stability
of highway traffic models  a much studied topic, in which
nevertheless we have found new ground! In carfollowing models,
perturbations always propagate backwards relative to platoons,
but since the vehicles drive forward in space, it is not clear
a priori whether "information" propagates forward or backwards
relative to the road. I will thus develop the concept of
"signal velocity" and show how it may be computed via
asymptotic methods. It turns out that the related concept of
"group velocity" is not quite the right thing to compute  and
in fact, the distinction between these two velocities dates
back to seminal work by Sommerfeld and Einstein at the
beginning of the 20th century.  This talk represents joint
work with Jonathan Ward (University of Limerick, Ireland) and
also ongoing conversations with Martin Treiber (TU Dresden).

Mathematical Criteria for Convective versus Absolute Instability



