Pid and model predictive control in a networked environment

Abstract

The networked environment presents many new challenges for the design of feedback control systems. Though networked control systems (NCSs) are subject to the performance-degrading effects of both discretization and time-varying, random delay, traditional control algorithms may still suffice depending on the plant, network, and performance criteria. This thesis investigates the implementation of two popular algorithms, PID control and model predictive control, in a networked environment. The stability properties of continuous-time systems with discrete control are analyzed under the simultaneous effects of discretization and loop delay for both algorithms using state-space models. We also investigate the use of play-back buffering to remove uncertainty in the delay and improve the performance of model predictive control in an NCS. The value of removing all delay uncertainty is studied with comparisons to unbuffered PID control using a bounded interval delay distribution. Additionally, optimal play-back buffer design for a heavy-tailed delay distribution is explored.