Planning Using Multiple Execution Architectures

Abstract

We discuss two techniques used by the RALPHMEA agent architecture to facilitate decision making in complex, real-time domains. Multiple execution architectures are four implementations of the agent function, a functionthat receives percepts from the environment as input and outputs an action choice. The four execution architectures are defined by the different knowledge types that each uses. Depending on the domain and agent capabilities, each execution architecture has different speed and correctness properties. Metalevel control of planning computes the value of information of planning to compare to the utility of executing the current default plan. Examples are presented from an autonomous, underwater vehicle domain.