A Multiple Rule Engine-Based Agent Control Architecture

This paper introduces a multiple rule engine agent architecture intended for prototyping heterogeneous agents. It also briefly describes how this same architecture might be used for concurrent and hierarchical goal processing. The proposed framework has an embedded main rule engine supporting goal processing as well as providing high level decision and control for the agent. It also possesses a mechanism for dynamically spawning rule engine-based agent behaviours (using separate rule engines) to process subgoals, and it maintains information/status communication channels between the main rule engine and the rule-based agent behaviours. The agent together with its agent behaviours are all realized in a similar manner. Each embedded rule engine has the ability to configure multiple agent behaviours either in parallel or hierarchically. In addition, this approach allows the construction of reuseable, rule-based agent capabilities that can easily be composed and layered to achieve a specific, application-oriented composite capability.