Knowledge-based and Data-driven Reasoning and Learning for Ad Hoc Teamwork

TL;DR

This paper introduces an architecture for ad hoc teamwork that combines knowledge-based reasoning with data-driven learning, enabling adaptable, transparent, and incremental modeling of other agents' behaviors in multi-agent collaboration.

Contribution

It proposes a novel hybrid architecture that integrates logical reasoning with incremental predictive models for improved ad hoc teamwork performance.

Findings

Supports adaptation to unforeseen changes

Enables incremental learning and model revision

Achieves better performance than baseline methods

Abstract

We present an architecture for ad hoc teamwork, which refers to collaboration in a team of agents without prior coordination. State of the art methods for this problem often include a data-driven component that uses a long history of prior observations to model the behaviour of other agents (or agent types) and to determine the ad hoc agent's behaviour. In many practical domains, it is challenging to find large training datasets, and necessary to understand and incrementally extend the existing models to account for changes in team composition or domain attributes. Our architecture combines the principles of knowledge-based and data-driven reasoning and learning. Specifically, we enable an ad hoc agent to perform non-monotonic logical reasoning with prior commonsense domain knowledge and incrementally-updated simple predictive models of other agents' behaviour. We use the benchmark simulated multi-agent collaboration domain Fort Attack to demonstrate that our architecture supports adaptation to unforeseen changes, incremental learning and revision of models of other agents' behaviour from limited samples, transparency in the ad hoc agent's decision making, and better performance than a data-driven baseline.