Networked control with stochastic scheduling

Kun Liu*, Emilia Fridman, Karl Henrik Johansson

*Corresponding author for this work

Research output: Contribution to journalArticlepeer-review

Abstract

This note develops the time-delay approach to networked control systems with scheduling protocols, variable delays and variable sampling intervals. The scheduling of sensor communication is defined by a stochastic protocol. Two classes of protocols are considered. The first one is defined by an independent and identically-distributed stochastic process. The activation probability of each sensor node for this protocol is a given constant, whereas it is assumed that collisions occur with a certain probability. The resulting closed-loop system is a stochastic impulsive system with delays both in the continuous dynamics and in the reset equations, where the system matrices have stochastic parameters with Bernoulli distributions. The second scheduling protocol is defined by a discrete-time Markov chain with a known transition probability matrix taking into account collisions. The resulting closed-loop system is a Markovian jump impulsive system with delays both in the continuous dynamics and in the reset equations. Sufficient conditions for exponential mean-square stability of the resulting closed-loop system are derived via a Lyapunov-Krasovskii-based method. The efficiency of the method is illustrated on an example of a batch reactor. It is demonstrated how the time-delay approach allows treating network-induced delays larger than the sampling intervals in the presence of collisions.

Original languageEnglish
Article number7063933
Pages (from-to)3071-3076
Number of pages6
JournalIEEE Transactions on Automatic Control
Volume60
Issue number11
DOIs
StatePublished - 1 Nov 2015

Keywords

  • Lyapunov functional
  • Networked control systems
  • Stochastic impulsive system
  • Stochastic protocols

Fingerprint

Dive into the research topics of 'Networked control with stochastic scheduling'. Together they form a unique fingerprint.

Cite this