RL-QN: A Reinforcement Learning Framework for Optimal Control of Queueing Systems

TL;DR

This paper introduces RL-QN, a reinforcement learning framework designed to optimize control policies in queueing networks, effectively minimizing average backlog despite large or unknown state spaces.

Contribution

RL-QN is a novel model-based RL algorithm that handles unbounded state spaces by focusing on finite subsets and applying stabilizing policies, achieving near-optimal performance.

Findings

RL-QN effectively minimizes average queue backlog.

Simulation results confirm RL-QN's superior performance.

The method applies to dynamic server allocation, routing, and switching.

Abstract

With the rapid advance of information technology, network systems have become increasingly complex and hence the underlying system dynamics are often unknown or difficult to characterize. Finding a good network control policy is of significant importance to achieve desirable network performance (e.g., high throughput or low delay). In this work, we consider using model-based reinforcement learning (RL) to learn the optimal control policy for queueing networks so that the average job delay (or equivalently the average queue backlog) is minimized. Traditional approaches in RL, however, cannot handle the unbounded state spaces of the network control problem. To overcome this difficulty, we propose a new algorithm, called Reinforcement Learning for Queueing Networks (RL-QN), which applies model-based RL methods over a finite subset of the state space, while applying a known stabilizing policy for the rest of the states. We establish that the average queue backlog under RL-QN with an appropriately constructed subset can be arbitrarily close to the optimal result. We evaluate RL-QN in dynamic server allocation, routing and switching problems. Simulation results show that RL-QN minimizes the average queue backlog effectively.