Thinking Together, Apart: Reasoning and Learning in Ad Hoc Teamwork

Hasra Dodampegama, Mohan Sridharan

Keywords: Knowledge representation, Non-monotonic logical reasoning, Ecological rationality, Ad hoc teamwork, Knowledge acquisition

TL;DR: We present an architecture that enables an hoc agent to coordinate with others using non-monotonic logical reasoning and learning based on decision heuristics.

Abstract: AI agents deployed to assist and collaborate with other agents often have to do so without prior coordination. Methods considered state of the art for such ad hoc teamwork pose it as a learning problem, using a large labeled dataset to model the action choices of other agents (or agent types) and determine the actions of the ad hoc agent. These methods lack transparency and make it difficult to rapidly revise existing knowledge in response to changes. Our architecture for ad hoc teamwork leverages the complementary strengths of knowledge-based and data-driven methods for reasoning and learning. For any given goal, we enables an ad hoc agent to determine its actions through non-monotonic logical reasoning with: (a) prior domain-specific commonsense knowledge; (b) models learned and revised rapidly to predict the behavior of other agents; and (c) anticipated abstract future goals based on generic knowledge of similar situations in an existing foundation model. The agent also processes natural language descriptions and observations of other agents' behavior, incrementally acquiring and revising knowledge in the form of objects, actions, and axioms that govern domain dynamics. We experimentally evaluate the capabilities of our architecture in VirtualHome, a realistic simulation environment.

Paper Track: Technical paper

Submission Number: 19