In this Section we commence with a brief description of the cell-based modeling paradigm (Section 8.2.2, Cellular automata), and then address the broader topic of agent-based modeling (ABM, Section 8.2.3, Agents and agent-based models, et seq) which now embraces many of the previously distinct class of CA models. Techniques of this kind have been described by some authors as geosimulation. As noted in Section 8.1, Introduction to Geocomputation, this is a general purpose term (introduced by Benenson and Torrens, 2004) used to describe the application of modern micro-level simulation tools to geospatial problems. Typically these problems have a well-defined (finite) spatial extent and have a dynamic context. Frequently a distinction is made between cell-based or zone-based spatial models and less formalized spatial models incorporating mobile ‘agents’. In practice there are close similarities between the two types of model. Toolsets designed to support both frameworks are widely available. For example MASON (a multi-agent simulation library discussed further in Section 8.2.11, Simulation/modeling (s/m) systems for agent-based modeling) can represent continuous, discrete, or hexagonal 2D, 3D, or network data, and any combination of these, with GIS support added in recent years. Another ABM toolkit with GIS support is GAMA, available from http://gama-platform.org/ (a French/Vietnamese collaborative project).
Application-specific tools (such as those used for urban and highway traffic simulation, e.g. MATSim and TRANSIMS) are not described here, although they have been discussed briefly in Section 7.2.1, Overview — network and locational analysis. A recent book on MatSIMS is recommended reading on this ABM toolset (Horni, Nagel and Axhausen, 2016). Also worthy of note is the planning support system UrbanSim, an open source land use modeling system written in Java that allows for the exploration of possible consequences of alternative transportation, land use and environmental policies. Various spatial datasets are used in the system such as environmental data (e.g. wetlands, slopes) and land parcel information. This information, along with employment and population data, is then used to explore future land use scenarios based on the actions and interactions of individual agents such as households and developers. For example, households have housing preferences and value access to workplaces and shopping opportunities which influence where they want to locate. UrbanSim has been applied to several metropolitan areas in the United States including Salt Lake City, Utah. A general overview of UrbanSim can be found in Waddell et al. (2008) while a more technical detail can be found in Waddell et al. (2003).