GenGPT: A Systematic Way to Generate Synthetic Goal-Plan Trees
Abstract: Deciding “what to do next” is a key problem for BDI agents with multiple goals, which is termed the intention progression problem (IPP). A number of approaches to solving the IPP have been proposed in the literature, however, their evaluations are all taken in different forms. The lack of standard benchmarks and testbeds for evaluating the IPP makes it difficult for researchers to contribute to this topic. To foster research around the IPP and BDI agents, this paper proposes a way to generate test cases in the form of goal-plan trees which can be used to represent the agent’s intentions in various agent languages and platforms.
Loading