Activity-based models (ABM) are gaining broader interest among public agencies across the world. However, resources and expertise needed for implementing a new modeling system or adapting an existing software program is a major and very valid concern for ABM adoption. Unlike trip-based models which typically conform to the 4-step paradigm, ABM vary greatly in, for example, modeling paradigm (e.g. econometric-based versus rule-based), overall system structure (e.g. with or without consideration of inter- and intra-household interaction), and model component design (e.g. choice of econometric structure). This leads to implementation challenges for developing reusable software artifacts and applications. To date, ABM have been implemented as customized applications using general-purpose programming languages, as opposed to macro and scripting languages provided by commercial transportation planning packages.
This paper contends that the activity-based modeling community can benefit greatly from a shared framework for the formal specification of ABM. It describes such a framework currently being developed at Parsons Brinckerhoff as part of a reusable activity-based modeling and simulation toolkit. The framework is primarily a language which has syntax to which domain specific symantics are added. This domain specific expression and modeling language (DSEML) provides the notations and constructs needed for encoding a calibrated ABM, from expressing the variables making up a utility function to defining choice alternatives, model parameters, component model structures, and potentially overall model flow. It exploits the grammar of a number of existing languages.
The benefits of the DSEML are many. First, integrating the DSEML -based specification to an ABM implementation effort promotes standardized documentation and enhanced system transparency. Second, the language is used in the ABM software’s input files to declare model components and their structures, thereby serving as an interface protocol between the simulation system and its user. Third, and most importantly, by raising the level of abstraction to represent the models as the available kinds of objects and data elements in the simulation system, the DSEML enables greater reuse of software artifacts, thereby reducing implementation costs.