The transport system is a crucial part of our daily life: It does impact where we work, where we spend our free time, but also where we live and how we are able to interact with friends, family and peers. It also takes an important role in making sure our supply chains remain stable and goods are available when we expect them to be so. To maintain and improve such a complex system and to adapt it to ever changing demand, transport planning has become an important field in research and engineering.
The agent-based transport simulation framework MATSim is a tool which can help practitioners with such planning tasks. Other than more aggregate transport models, MATSim simulates each individual traveller in the system throughout one typical day. By interaction of agents, congestion emerges and traffic patterns arise. Transport planners can then introduce changes into the system and assess the impact of these actions. Because of MATSim’s granularity, detailed analyses of regions, towns, roads, train tracks and even individual travellers can be performed.
MATSim has been under development for more than ten years as a collaboration between ETH Zurich and TU Berlin. The focus has been mainly on conceptual and methodological improvement of the software. At the same time a large international user base has grown, which is today pushing the system forward to increasingly ambitious applications. Since 2017 the Swiss Federal Railways are assessing the applicability of a MATSim simulation of Switzerland for their service planning. Therefore, in the framework of the ETH Mobility Initiative, the project partners aim at substantially improving the simulation runtime of MATSim to enable its use as an element in a realistic planning process.
To that end, the Swiss Federal Railways make their planning model for Switzerland available to the Systems Group and the Institute for Transport Planning and Systems at ETH Zurich. While the former will use their expertise in High Performance Computing to speed up the software code and assess pathways for running the simulation in a distributed environment on a computation cluster, the latter will further advance the methods used and explore ways of improving stability of simulation results and convergence. Ultimately, a speed-up of the framework of up to 100x is seeked in the “High Performant Mobility Simulation” project.
Dr. Michel Müller, Sebastian Hörl
Institute for Transport Planning and Systems, ETH Zürich
07.2018 (36 months)