Observations weighted controllers for linear stochastic systems (Q1081581)

This paper derives a controller for self-tuning applications where significant coloured noise or output disturbances are present. The observation weighted (OW) controller derived here is based on the minimization of a cost function including an unconditional (LQG type of) expectation operator. The plant description is more general than that usually considered in self-tuning problems, since it allows for the presence of both input and output disturbances. The solution for the controller is relatively simple to implement. By selecting the cost function weightings appropriately by OW controller can be used in a pole placement mode. The advantage of deriving a pole placement design via this optimal approach is that the solution will take account of the presence of both input and output disturbances. Whether this is a reasonable strategy to choose the cost weights will depend upon the application of interest.