Introduction
In
order to explore and evaluate future potential scenarios that involve new
policies, infrastructure changes or even minor operational logic changes,
simulation models are often the most reliable option. On the other hand, the
complexity of all relevant interactions in a simulation model demands
simplifications that often compromise the validity of the results. For example,
in mesoscopic traffic simulation models, vehicle movement can be determined by
representations such as speed-density relationship functions, with parameters
calibrated a priori. While this provides reliable results under habitual
circumstances, under new scenarios such as incidents or infrastructure changes,
the demand or supply model assumptions might have changed. The use of decoupled
models to solve this (e.g. using a microscopic model to obtain new parameters
where/when needed) can be a challenging solution since it demands full
consistency between models (e.g. mesoscopic and microscopic) and is often
difficult to implement in practice. From this perspective, a fully integrated
simulation model, that considers macro, meso and microscopic levels, is an
ongoing challenge with significant impact for future research and practice. There
has been some progress in developing integrated frameowrks, however, most of
the efforts have shown loosely coupled integration between demand (usually
activity-based model(ABM)) and supply (dynamic traffic assignment(DTA)) models.
SimTRAVEL (1) shown a step ahead by
inserting landuse component within the integrated framework of ABM + DTA. Also,
POLARIS (2) has claim to provide more
flexible agent based platform to integrate separate models together but so far
has been applied only to a level of integration that can be classified as
ABM+DTA. Development towards fully integrated platform that integrates
long-term, mid-term and short term model together is scarce, primarily due to
its challenging nature.
This paper present a fully integrated agent based platform
named as SimMoblity which integrates
various mobility-sensitive behavioural models within a multi-scale simulation
platform that considers land-use, transportation and communication
interactions. It focuses on impacts on transportation networks, intelligent
transportation services and vehicular emissions, thereby enabling the
simulation of a portfolio of technology, policy and investment options under
alternative future scenarios. SimMobility incorporates three different
sub-models:
- Short-term(ST) simulator - The time step can be a
fraction of a second and agent decisions include lane changing, braking,
accelerating, gap acceptance, but also route choice. SimMobility short-term
model is a traffic micro-simulator (e.g. (3,
4)), extended with a communications simulator as well as pedestrians and
public transport.
- Mid-term(MT) simulator - The time step is in the range of
seconds to minutes and agent decisions include route choice, mode choice,
activity pattern and its (re)scheduling, departure time choice. SimMobility
mid-term is a mesoscopic simulator (e.g. (5,
6)), designed for activity-based modeling, with explicit pre-day and
within-day behavior including re-routing and re-scheduling, and multiple
transport modes.
- Long-term(LT) simulator - The time step is in the range
of days to months to years, and agent decisions include house location choice,
job location choice, land development, car ownership. It is a land-use and
transport (LUT) simulator (e.g. (7,8)),
with a market transaction bidding model.
This paper introduces the full SimMobility system, with
focus on the innovative contributions that span across all levels. We emphasize
the benefits and feasibility of a fully integrated approach, which relies, by
design, on the acitvity-based modeling paradigm, with implications for all
levels. The paper then disucsses the SimMobility through a case study with
autonomous mobility which may have impacts on long-term, mid-term and short-tem
levels, and is specifically designed to showcase advantages of integration
across all thee levels .