Using model predictive control for real-time control over the Internet

The performances of two MPC (model predictive control) techniques, move-suppressed and shifted, are studied under normally distributed random communication delays to test the potential of MPC as a control method over the Internet. A conventional PID is also implemented under the same delay conditions. All controls are run in a time-scheduling scheme. The results show the failure of the PID to provide stable control even under small delays while the two MPC are able to accommodate fairly large delays and provide stable control that gets slower with larger delays. It was also found that the very same MPC parameters (move-suppression and shift factors) used to reduce the inherent ill-conditioning of the MPC dynamic matrix, play the main role in adapting to randomness in the communication.