Large-Scale Agent-Based Transport Simulation in Shanghai, China

Abstract

The activity-based model system is being coined as the next-generation demand-forecasting model. The agent-based transport simulation toolkit MATSIM is a fully integrated system that models decisions from the long term to the short term, and these decisions in MATSIM are activity-based models. This paper describes the application of MATSIM in a large-scale multiagent-based transport simulation for Shanghai, China. First, algorithms for integrating new data in Shanghai with MATSIM inputs such as synthetic population, facilities, and network are separately designed according to data characteristics. Then activity-based modeling is introduced to generate population plans, and activity replanning is employed to learn the better travel plans; a utility-based approach is used to model scoring for a plan. Finally, a full MATSIM-based simulation platform for the Shanghai scenario is built in detail. The scenario contains 200,000 synthetic persons simulated on a network with 50,000 links. The relaxed state of the simulation system is reached after 100 iterations of replanning procedures, and the mode choice, route choice, and activity time allocation modules are used to optimize agents’ activity plans. The feasibility of transport simulation in Shanghai by MATSIM is validated against the mode split and the observed counts. Extensive simulation results for the designed Shanghai simulation scenarios indicate that most of the observed counts match quite well with the traffic simulation volumes and demonstrate the potential of MATSIM for large-scale dynamic transport simulation.

Social media and mobility landscape: Uncovering spatial patterns of urban human mobility with multi source data

“…Population synthesis is a widely-adopted technique for creating a full population microdata based on limited number of observations, to enable analyzing estimates of variables at different spatial scales. It was first developed in the field of economics in the 1960s, and has since been introduced to the fields of geography and social sciences (Hermes and Poulsen, 2012;Ma et al, 2014), for example, to analyze social policy and population changes (Haase et al, 2010), transportation (Beckman et al, 1996;Zhang et al, 2013;Ma et al, 2014), building energy consumption (Subbiah et al, 2013) and so on. In the transport-related field, the pioneering attempt of population synthesis goes back to 1996, when Beckman et al (1996) applied Iterative Proportional Fitting (IPF) to create a baseline synthetic population of individuals and households, so as to estimate future travel demand.…”

Automated identification and characterization of parcels with OpenStreetMap and points of interest

“…Preliminary results suggest that inter alia (1) OSM data are more dense and applicable in major cities, although the overall data coverage may be poor; (2) data completeness remains the major issue (Goodchild, 2008); (3) OSM data in China have been continuously improved. In line with observations, OSM data are gradually being used to understand Chinese cities (see, for example, Leitte et al., 2012; Liu et al., 2012; Zhang et al., 2013).…”

An improvement in MATSim computing time for large-scale travel behaviour microsimulation

“…Researchers primarily from ETH Zurich and TU Berlin have been working on MATSim for more than a decade (Horni et al 2016), which also provides an integrated framework for DTA and Agent-based Modelling. Compared with other integrated frameworks, MATSim is one of the most popular, as evident from its worldwide applications (MATSim 2015), including China (Zhang et al 2013;Zhuge et al 2014), Switzerland (Bekhor et al 2010), Belgium (Röder et al 2013), Germany (Neumann et al 2012), Canada (Gao et al 2010), Singapore (Axhausen 2013), and South Africa (Neumann et al 2015).…”

“…Large SD has become a common issue in similar simulation work using MATSim. For instance, when Zhang et al (2013) applied MATSim to simulate the travel behaviour in Shanghai and the simulation outcomes (their Figure 9) indicated that the gap between the simulated and observed traffic flow could be quite large for certain individual links.…”