AI agents in practical domains often have to cooperate with other agents without prior coordination. State of the art approaches for such ad hoc teamwork pose this task as a learning problem, using a large 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. We present an architecture for ad hoc teamwork that leverages the complementary strengths of knowledge-based and data-driven methods for reasoning and learning. For any given goal, the ad hoc agent determines 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 our architecture’s capabilities in VirtualHome, a realistic simulation environment.

State of the art methods for ad hoc teamwork, i.e., for collaboration without prior coordination, often use a long history of prior observations to model the behavior of other agents (or agent types) and to determine the ad hoc agent's behavior. In many practical domains, it is difficult to obtain large training datasets, and necessary to quickly revise the existing models to account for changes in team composition or domain attributes. Our architecture builds on the principles of step-wise refinement and ecological rationality to enable an ad hoc agent to perform non-monotonic logical reasoning with prior commonsense domain knowledge and models learned rapidly from limited examples to predict the behavior of other agents. In the simulated multiagent collaboration domain Fort Attack, we experimentally demonstrate that our architecture enables an ad hoc agent to adapt to changes in the behavior of other agents, and provides enhanced transparency and better performance than a state of the art data-driven baseline.

Abstract Ad hoc teamwork (AHT) refers to the problem of enabling an agent to collaborate with teammates without prior coordination. State of the art methods in AHT are data-driven, using a large labeled dataset of prior observations to model the behavior of other agent types and to determine the ad hoc agent’s behavior. These methods are computationally expensive, lack transparency, and make it difficult to adapt to previously unseen changes. Our recent work introduced an architecture that determined an ad hoc agent’s behavior based on non-monotonic logical reasoning with prior commonsense domain knowledge and models learned from limited examples to predict the behavior of other agents. This paper describes KAT, a knowledge-driven architecture for AHT that substantially expands our prior architecture’s capabilities to support: (a) online selection, adaptation, and learning of the behavior prediction models; and (b) collaboration with teammates in the presence of partial observability and limited communication. We illustrate and experimentally evaluate KAT’s capabilities in two simulated benchmark domains for multiagent collaboration: Fort Attack and Half Field Offense. We show that KAT’s performance is better than a purely knowledge-driven baseline, and comparable with or better than a state of the art data-driven baseline, particularly in the presence of limited training data, partial observability, and changes in team composition.