Feedback Scheduling of Model Predictive Controllers

The paper presents some preliminaryresults on dynamic scheduling of model predictive controllers (MPCs).In an MPC, the control signal is obtained by on-line optimization of acost function, and the MPC task may experience very large variationsin execution time from sample to sample. Unique to this application,the cost function offers an explicit, on-line quality-of-servicemeasure for the task. Based on this insight, a feedback schedulingstrategy for multiple MPCs is proposed, where the scheduler allocatesCPU time to the tasks according to the current values of the costfunctions. Since the MPC algorithm is iterative, the feedbackscheduler may also abort a task prematurely to avoid excessiveinput-output latency. A case study is presented, where the newapproach is compared to conventional fixed-priority andearliest-deadline-first scheduling. General problems related to thereal-time implementation of MPCs are also discussed.