Abstract:
The modelling of private vehicle and freight tra c using agent-based simulation (ABS) has
been very successful in South Africa, but transit, or public passenger transport, has compara-
tively lagged behind. This research project implemented a large-scale multi-modal Multi-Agent
Transport Simulation Toolkit (MATSim) model, which includes transit, for the Nelson Mandela
Bay Metropole (NMBM). To achieve this, the MATSim simulation environment is investigated
and available General Transit Feed Speci cation (GTFS) data for the region is matched to
the network and transformed into a MATSim transit schedule. A comparison of planned and
simulated transit leg times indicate that for short legs, simulated travel times are higher than
planned, while for longer legs, simulated travel times are shorter than planned. These values
are expected to improve as transit data accuracy and model calibration are re ned. In addition
to the comparison of planned and simulated values, several analyses provide detailed results for
and insight into the NMBM transit system. This project, as the rst large-scale agent-based
transit simulation in South Africa, represents the rst step in a process towards providing mod-
els which can accurately capture the intricacies of our public transport environment. Several
research projects are currently under way which aim to model minibus taxis - only once this
integral part of our system is included will agent-based transit simulation be considered as truly
representative, but even now it a ords the opportunity to gain unparalleled insight into current
and future transit systems.
