Stability Analysis for Model Based Predictive Control System with Network Induced Delay and External Disturbance

Abstract

Using a packet switching network is an effective way to construct systems that use multi-actuators and sensors; however, the structural characteristics of a packet switching network, i.e., packet loss and packet transient time delay, affect the control performance of the systems that use a packet switching network. The model based predictive control scheme compensates well for the effect of the packet transient time delay, and the effect of packet loss as long as the resulting maximum delay is known; however, existing work has not considered external disturbance. External disturbance represents the unexpected control input factors on the operation environment. This thesis analyzes the effects of external disturbance on the model based predictive control system. In conventional feedback control systems, disturbance does not affect system stability; however, in systems with model based predictive control method, external disturbance may destabilize the control system. Existing work on model based predictive control does not properly analyze the above mentioned phenomenon. We use a lifting technique with current states, delayed states, and prediction error, in order to analyze closed loop system stability and performance with external disturbance. We show that the system matrix with lifted states define the closed loop stability. The results are validated experimentally on a Ball Balancer control system with time delay. The control system with a controller designed using the stability analysis of this thesis, exhibits stable behavior despite external disturbance and compensates for the performance loss due to time delay. ⓒ 2016 DGIST