| dc.description | Online adaptation of the Simplified Predictive Control (SPC) from [14] to be used via IP networks is developed. No knowledge of network delay distributions, delay bounds or stationarity is assumed. Five protocols are proposed for scheduling the control tasks and communicating the control data and its usage in the additive feedback correction of the predictions. A generic, semi-discrete, analytical study of SPC is developed for each protocol. Stability analysis of the resulting delay differential equations (DDEs) allow for an understanding of the twin effects of measurement timestamping and assumptions regarding the timing of move application. To enhance the stability and robustness to random delay of the best performing protocol, a two-parameter variant of the SPC is formulated and Fuzzy-sensing of the network delay state is used to Fuzzy-schedule the two parameters of the SPC and adapt to future delay. A PI-SPC equivalence is developed in both cases of conventional and networked controls. It allows commonly-used PI control to provide equivalent control to SPC via a move-to-move PI parameter adaptation. All control methods are implemented as an Application layer protocol in the IP stack with the User Datagram Protocol (UDP) functioning as a Transport protocol. The implementation is based on the ns-2 simulator which is modified here to include a controller and necessary plant agents. | en_US |
