Goal Reasoning with Informative Expectations

In complex and dynamic scenarios, autonomous vehicles often need to intelligently adapt their behavior to unexpected changes in the their environment. Goal Reasoning provides a methodology for autonomous agents to deliberate and adapt their goals to more intelligently react to changing conditions. This paper implements a Goal Reasoning system based on the Goal Lifecycle, and grounds the implementation in the information measures and expectations used by the vehicles to asses their performance. The implemented system, termed Goal Reasoning with Information Measures (GRIM), is demonstrated using a disaster relief scenario in which a small team of vehicles is tasked with surveying a pre-defined set of geographical regions. This demonstration shows how area search goals can be progressively refined, and how they can be adapted to resolve problems encountered by the vehicles