Event-Triggered Stochastic Model Predictive Control for Constrained Queueing Networks

Li Deng

doi:10.1109/TNSE.2023.3330437

Return

Event-Triggered Stochastic Model Predictive Control for Constrained Queueing Networks

2024-03-01

PRE

邓力 (Li Deng) *

Zhan Shu

Tongwen Chen

DOI：10.1109/TNSE.2023.3330437

Original

Original request for help

Save

Abstract

En 中文

An event-triggered stochastic model predictive control (MPC) approach is proposed for the scheduling problem of constrained queueing networks with a dynamic topology. A discrete-time Markov chain (DTMC) in combination with a Bernoulli trial is used to model the time-varying routing of queueing networks. The constituency and positiveness constraints on queue lengths together with the dynamic topology and the stochasticity in packet arrival are incorporated into a stochastic MPC optimization problem. An event-triggered scheme with adaptive event checking involving an estimated waiting horizon is designed to trigger the solution of the optimization problem when necessary, leading to reduced computational burden and improved utilization of communication resources. The stability is analyzed by the relation between the inter-execution time and objective function. The proposed approach is applied to two queueing networks to show its effectiveness.

Keywords:

Optimization

Topology

Network topology

Routing

Monitoring

Markov processes

Predictive models

Adaptive checking strategy

constrained queueing networks

event-triggered stochastic MPC

Markov chains