Generating problem episodes for a multi-agent syste: The TAEMS Grammar Generator

Empirical studies are an important method of determining the roles of different algorithms in multi-agent systems research. In order to run multi-agent experiments, one needs a method of generating problem instances for the agents to solve. These problem instances can then be converted into TAEMS task structures, which are hierarchical abstractions of the problem solving process that describe alternative ways of accomplishing the desired goal. Most real application domains constrain the morphology of the task structures, therefore in order to generate appropriate problem instances, a model of the environment that the multi-agent system is operating in is needed. We present the TAEMS grammar generator, a environment modeling tool capable of capturing the constraints present in the domain task structures, along with the uncertain and dynamic nature of most real world environments.