Centralized Training with Hybrid Execution in Multi-Agent Reinforcement Learning

TL;DR

This paper introduces hybrid execution in multi-agent reinforcement learning, enabling agents to adapt to varying communication levels at runtime using a centralized predictive model, improving cooperation under partial observability.

Contribution

The paper proposes hybrid-POMDPs to model variable communication levels and introduces MARO, a centralized auto-regressive model that estimates missing observations during execution.

Findings

MARO outperforms baselines in standard benchmarks.

Agents effectively cooperate with faulty communication.

Hybrid execution improves robustness in partial observability.

Abstract

We introduce hybrid execution in multi-agent reinforcement learning (MARL), a new paradigm in which agents aim to successfully complete cooperative tasks with arbitrary communication levels at execution time by taking advantage of information-sharing among the agents. Under hybrid execution, the communication level can range from a setting in which no communication is allowed between agents (fully decentralized), to a setting featuring full communication (fully centralized), but the agents do not know beforehand which communication level they will encounter at execution time. To formalize our setting, we define a new class of multi-agent partially observable Markov decision processes (POMDPs) that we name hybrid-POMDPs, which explicitly model a communication process between the agents. We contribute MARO, an approach that makes use of an auto-regressive predictive model, trained in a centralized manner, to estimate missing agents' observations at execution time. We evaluate MARO on standard scenarios and extensions of previous benchmarks tailored to emphasize the negative impact of partial observability in MARL. Experimental results show that our method consistently outperforms relevant baselines, allowing agents to act with faulty communication while successfully exploiting shared information.