Agent Probing Interaction Policies

TL;DR

This paper explores the use of probing policies to identify agent types in multi-agent reinforcement learning environments, addressing non-stationarity by extending an existing probing framework.

Contribution

It introduces an extension of the Environmental Probing Interaction Policy framework for multi-agent settings, assuming stationary policies of other agents.

Findings

Probing policies improve agent type identification.

Extension of probing framework to multi-agent environments.

Addresses non-stationarity in reinforcement learning systems.

Abstract

Reinforcement learning in a multi agent system is difficult because these systems are inherently non-stationary in nature. In such a case, identifying the type of the opposite agent is crucial and can help us address this non-stationary environment. We have investigated if we can employ some probing policies which help us better identify the type of the other agent in the environment. We've made a simplifying assumption that the other agent has a stationary policy that our probing policy is trying to approximate. Our work extends Environmental Probing Interaction Policy framework to handle multi agent environments.