DEFINING AND VISUALIZING MODELS OF URBAN TRAFFIC
Shannon Borho, Jan Pittner, Gabriel Wainer
Department of Systems and Computer Engineering
Carleton University
4456 Mackenzie Building. 1125 Colonel By Drive
Ottawa, ON. K1S 5B6. CANADA
gwainer@sce.carleton.ca

ABSTRACT
ATLAS is a modeling language that permits defining a
static view of a city section for simulating traffic in an area.
The models are formally specified, avoiding a high number of
errors in the application, thus reducing the problem solving
time. The system required the manual generation of ATLAS
files, a tedious process that did not lend itself for rapid changes
to the system input. The output of the system also suffered from
a non-user friendly interface. The solutions to these problems
were addressed in two parts: a front-end system allowing the
user to draw city sections (and then parse the drawing to create
a valid ATLAS file), and an output subsystem permitting showing
cars with realistic 3D graphics.
1. INTRODUCTION
Urban traffic analysis and control is a problem of such a
complexity that it is difficult to be analyzed with traditional
analytical methods. Modeling and simulation techniques, instead,
have shown a certain degree of success, and they have
been gaining popularity as an analysis tool. Simulation permits
studying particular problems using virtual experimentation.
We have developed a toolkit for modeling and simulation
of traffic in urban centers. This project followed a rigorous approach
that we introduce here. The first stage was devoted to
define and validate a high level specification language representing
city sections [1]. This language, called ATLAS (Advanced
Traffic LAnguage Specifications) focuses on the detailed
specification of traffic behavior. The models are
represented as cell spaces, allowing elaborate study of traffic
flow according to the shape of a city section and its transit attributes.
A static view of the city section can be easily described,
including definitions for traffic signs, traffic lights, etc.
A modeler can concentrate in the problem to solve, instead of
being in charge of defining a complex simulation.
The constructions defined in this language are mapped
into DEVS [2] and Cell-DEVS models [3]. DEVS provides
high performance for discrete-event systems simulation [4].
Similar results were obtained for Cell-DEVS models [5]. It also
provides a formal framework that can be used to validate and
verify the models. This approach permits us to reuse the models
created and integrate with others using different formalisms
(for instance, using Petri Nets or Finite State Machines to specify
the behavior of traffic lights or railway controllers).
A real system modeled using the DEVS formalism can be
described as being composed of several submodels. Each of
them can be behavioral (atomic) or structural (coupled). A
DEVS atomic model is described as:
M = < X, S, Y, dint, dext, l, D >
Here, X is the input events set, S is the state set, and Y is
the output events set. We also use four functions: dint manages
internal transitions, dext external transitions, l the outputs, and
D, the lifetime of a state. The interface is composed of input
and output ports to communicate with other models. Each port
is defined as a pair, including a port name and its type. The input
external events (those coming from other models) are received
in input ports.
A DEVS coupled model is defined as:
CM = < I, X, Y, D, {Mi}, {Ii}, {Zij} >
Here, I is the model interface, X is the set of input events,
and Y is the set of output events. D is an index of components,
and for each i ΠD, Mi is a basic DEVS model (atomic or coupled).
Ii is the set of influencees of model i. For each j ΠIi, Zij
is the i to j translation function. Each coupled model consists of
a set of basic models connected through the input/output ports.
The influencees of a model will determine to which models one
send the outputs. The translation function is in charge of translating
outputs of a model into inputs for the others. To do so, an
index of influencees is created for each model (Ii). For every j
in this index, outputs of the model Mi are connected to inputs in
the model Mj.
The Cell-DEVS formalism was proposed as an extension
to DEVS permitting to describe cellular models. Cell-DEVS
allows the definition of complex cellular models that can be
integrated with other DEVS. Here, each cell of a space is defined
as an atomic DEVS with explicit timing delays. Transport
and inertial delays define the timing behaviors of each cell in an
explicit and simple fashion. A transport delay allows us to
model a variable response time for each cell. Instead, inertial
delays are preemptive: a scheduled event is executed only if the
delay is consumed.
Figure 1: Informal Definition of Cell-DEVS.
Cell-DEVS atomic models can be formally specified as:
.....