Goal Reasoning for an Autonomous Squad Member

Autonomous agents are beginning to play larger roles within team-oriented tasks and missions in various domains. Many reasoning agents are designed to perform in or assist with a single, static goal or task within an environment. Our aim is to design and develop an autonomous squad member that assists a squad with conducting a surveillance mission by identifying and reacting dynamically to changing situations and goals. We present a goal reasoning system for this agent that integrates natural language processing, explanation generation, and plan recognition components to recognize these changing situations and the squads responses to them. Our system uses goal selection and plan generation to respond to such changes. We describe the architecture we use to integrate these components and provide a case study that demonstrates how they work together to make a robust and adaptive autonomous agent.