Goal Reasoning to Coordinate Robotic Teams for Disaster Relief

Goal reasoning is a process by which actors deliberate to dynamically select the goals they pursue, often in response to notable events. Building on previous work, we clarify and define the Goal Reasoning Problem, which incorporates a Goal Lifecycle with refinement strategies to transition goals in the lifecycle. We show how the Goal Lifecycle can model online planning, replanning, and plan repair as instantiations of Goal Reasoning while further allowing an actor to select its goals. The Goal Reasoning Problem can be solved through goal refinement, where constraints introduced by the refinement strategies shape the solutions for successive iterations. We have developed a prototype implementation, called the Situated Decision Process, which applies goal refinement to coordinate a team of autonomous vehicles to gather information soon after a natural disaster strikes. We outline several disaster relief scenarios that progressively require more sophisticated responses. Finally, we demonstrate the prototype Situated Decision Process on the simplest of our scenarios, showing the merits of applying goal refinement to disaster relief.