IMPORTANCE OF REALISTIC MOBILITY MODELS FOR VANET NETWORK SIMULATION
Description of the project thesis at Fraunhofer ITWM
1. Description of the project thesis at Fraunhofer ITWM
Michael Speckert, Fraunhofer ITWM
Kaiserslautern, February 21, 2014
At ITWM, a project called VMC is currently executed which is partially funded by the
truck manufacturers DAF, Daimler, MAN, Scania and Volvo. An implementation of the
corresponding methods and the development of a database for environmental data are
supported by a joint project with the industry partners DAF, Daimler, MAN, Scania, and
Volvo.
An important part is the calculation of a speed profile given a route, some simple
vehicle characteristics and some parameters describing the driver behaviour.
A basic implementation exists, which has to be checked, validated and further
improved. Especially the resulting accelerations are of importance and need to be
analysed carefully. The approach uses a fixed time stepping scheme and is controlled by
a set of parameters. The influence of these parameters on the results needs to be
analysed as well as the interdependency of the parameters. In addition some
enhancements regarding the underlying simple model for longitudinal dynamics are
intended.
Some details of the speed profile calculation are sketched below.
The Speed Profile Calculation Approach
a) Tracks and routes
A track is a single, connected sequence of pieces of roads.
A route object is a collection of one or more tracks not necessarily connected to each
other.
Speed profiles are calculated for tracks. The results are collected for describing
routes.
b) Constraints along the route
A route allows access to all information (tags) belonging to the corresponding OSM
nodes/ways, as each section is part of a single way. This is important for getting all
information necessary for speed profile calculation such as speed limits or traffic
signs.
c) Driver and vehicle model
The speed profile generation makes use of a driver and a vehicle model, which
describes how the vehicle accelerates or brakes, and how it drives through a curve.
Therefore, we have to know the longitudinal acceleration and deceleration limits,
the maximum allowed lateral acceleration, and the influence of slope on the
vehicle's acceleration capabilities. For the calculation of speed profiles, the minimum
of the acceleration limits describing on the one hand the driver behaviour and on
the other hand the vehicle dynamics has to be chosen.
A further driver property is to deviate from the given speed limits and to drive
slower or faster. This is modelled by randomizing the speed limit (due to legislation)
with a Gaussian AR(1)-process with mean value 1, as defined in modelling the local
2. traffic density [1]. Two parameters describing different types of drivers are the
variance and the auto-correlation coefficient (η).
The corresponding parameters are defined via configuration files (xml-format).
Concerning the vehicle model, we define an xml-format capable of handling the
vehicle parameters needed for the speed profile generation (mass and acceleration
properties) and the parameters needed for the vertical virtual measurement analysis
(wheel mass, sprung mass, stiffness and damping values) described later. The first set
of parameters describes the horizontal part of the model, the latter the vertical part
of the model.
d) Traffic model
The speed profile generation needs information about the expected traffic during
driving. The speed profile generation relies on the spatial traffic density (number of
vehicles per km) along the route, which is usually not contained in the VMC
database.
We obtain the spatial traffic density from a temporal traffic density (e.g. number of
vehicles per day, which might be contained in the database) together with a speed
assumption and an assumption about how to map the daily average to specific
hours of a day.
If there is no traffic information available, the user must define some corresponding
assumptions, for instance, via a table of segments and its spatial traffic density.
The density values obtained in one of the ways described above are global traffic
densities, i.e. applying to a (long) road segment. For a more realistic modelling the
global density is randomized with a Gaussian AR(1)-process with mean value 1, see
[1]. The stochastic parameters defining this process can either be defined by the user
or taken from the default parameterization, which we will implement in a
configuration file.