| description |
In the research field of microscopic simulation models for
pedestrian mobility substantial pro-gress was achieved in recent
years. The ongoing strong growth of computational power to-gether
with new approaches founding on cellular automata allows the
microscopic simulation of large-scale scenarios like the evacuation
of whole soccer stadiums, as self-organization can be initiated by
using a relatively simple rule set. Recent developments at the
Institute for Road and Transportation Science will now ease the
simulation of everyday situations with multiple source-destination
relations as they occur e.g. in city centers, airport terminals and
public transport interchange points. Regardless of its level of
detail, each mobility simulation model is in need of comprehensive
empirical data for validation and calibration purposes. Thereby
macroscopic mobility models, e.g. traffic assignment models, only
need cross sectional data like vehicles per time seg-ment for
instance. Concerning pedestrian mobility on the large scale, only
macroscopic sur-veys like cordon counting in a public transport
system or traveler flow counting in an airport terminal have been
carried out up to now. However, the empirical data resulting from
these surveys is not suitable for calibrating and validating
microscopic simulation models for pedes-trian traffic, because they
only deliver aggregated traffic flow data. Therefore it was
neces-sary to develop a new survey method which allows the
noninvasive recording and analysis of pedestrian traces. In
collaboration with the Institute of Transport Research of the German
Aerospace Center (DLR) a two-day video survey at the Campus
Vaihingen of Universität Stuttgart was carried out. By a network of
four wide-angle cameras placed at strategic viewpoints, a central
area of the campus of about 20,000 sqm incorporating the commuter
railway station and the sur-rounding buildings was recorded at a
rate of five frames per second with a spatial resolution of 1024x786
pixels of each camera. In order to analyze the collected imagery of
some giga-bytes of raw data, solutions for several challenges had to
be found. First, suitable tie and control points for georeferencing
of the respective cameras had to be identified. This is re-quired to
transform the pedestrians trajectories from the respective images to
a common reference system. Second, methods for deriving pedestrian
trajectories from consecutive frames as well as the distinction
between pedestrians movements and the movement of all other objects
had to be improved. Third, unique tags had to be assigned to the
observed pe-destrians to trace them over the whole monitored area
while they are tracked by the different cameras. The resulting new
and unique data set not only consists of pedestrian traces with
decided timestamps at each waypoint, it also includes the resulting
source-destination matrix on which the timestamp for each pedestrian
allows many kinds of statistical operations, e.g. the computation of
the average walking time and standard deviation on different
source-destination relations. The availability of these real traces
will finally enable extensive and high-quality calibration and
validation operations for microscopic simulation models. The dataset
may even serve as input information for simulation in the research
field of communi-cation networks, in particular, in mobile ad hoc
networks. In the near future, the detailed ex-amination of many
other problems, such as counter flow, crossing behavior, and path
build-ing phenomena based on real microscopic mobility data will
become possible.
|