A Simulation Framework for Heterogeneous Agents

We introduce a generic simulation framework suitable for agent-based simulations featuring the support of heterogeneous agents, hierarchical scheduling and flexible specification of design parameters. One key aspect of this framework is the design specification: we use an XML-based format which is simple-structured yet still enables the design of flexible models. Another issue in agent-based simulations, especially when ready-made components are used, is the heterogeneity arising from both the agents’ implementations and the underlying platforms. To tackle these kind of obstacles, we introduce a wrapper technique for mapping the functionality of agents living in an interpreter-based environment to a standardized JAVA interface, thus facilitating the task for any control mechanism (like a simulation manager) because it has to handle only one set of commands for all agents involved. Again, this mapping is made by an XML-based definition format. We demonstrate the technique by applying it to a simple sample simulation of two mass marketing firms operating in an artificial consumer environment. 1 Overview: The Simulation Concept Simulations are often implemented by using an object-oriented style of programming, allowing for distributed objects (over a cluster of workstations or even the Internet). In the following, we consider the example of an agent-based economic simulation: one can think of each economic entity as an agent, e.g., firms, (groups of) consumers, investors, markets, etc., interacting with each other. A typical simulation combines several agents, defines their relationships and observes the resulting interactions between them over time. After the simulation design has been defined (see, e.g., Richter & März, 2000), running a simulation usually amounts to writing a control program in one’s favorite programming language, named the simulation manager (see below), coordinating a set of previously implemented, autonomous agents. One might wish that all agents should provide standardized interfaces such that they automatically have the same bindings and thus can be used for simulations as modularized components. General mechanisms for providing standardized interfaces (like CORBA) do exist, but usually require advanced programming skills to be used. We therefore offer an easy-to-use mechanism enabling the integration of data analysis environments like MATLAB, Octave or R, as they offer convenient ways of analyzing simulation results and are also (typically) used for implementing objects and methods. Another key issue is the variation of parameters in controlled experiments. Furthermore, we need a scheduling scheme determining the order of invocation within a single experiment (design) and the number of runs (periods) per design. Consider a simple example involving two competing firms, named "Firm A" and "Firm B", respectively, operating on a consumer market (see Figure 1). Each firm could be modularized itself, having agents responsible for marketing, production and finance. Market coordination and clearing may be performed by a consumer market agent, which models a (disaggregated) consumer population. In addition, a global environment, representing the common knowledge of all agents, is typically involved: this environment may be stored, e.g., in an SQL database, thus solving problems arising from simultaneous access by different agents (like blocking etc.), or managed by an information broker similar to the one described in Wilson et al. (2000)—but these mechanisms are highly specific to the simulation design. After a brief review of related work, we therefore restrict the presentation of our software1 to: • The specification of the simulation settings in XML for a generic simulation manager, supporting multiple design specification, and • The “normalization” of agent interfaces via a wrapping technique, thus allowing the simulation manager to treat all agents the same way. We complement the technical presentation by an application to a simulation of two mass marketing firms operating in an artificial consumer environment. available at: http://elrond.ci.tuwien.ac.at/software/simenv-1.0.tgz