Algorithm for PID controller tuning using LQG cost minimization

The optimisation of time domain cost functions to generate tuning rules has been well explored. However, there have been very few extensions of this procedure to an optimal PID trade-off from a formal time domain or frequency domain LQG or LQ cost minimisation problem. In this paper an LQG cost minimisation problem is defined and the PID controller coefficients which minimise the index are required. The advantages of the approach are that direct cost function optimisation is performed rather than simply matching a PID controller to an LQG solution. This enables the LQG underlying problem to be used more directly for the design process. A key idea is to choose the LQG cost function specification allowing for the constraints of the PID structure. The result is a PID design approach which is very closely integrated with the underlying optimisation problem. In the paper the algorithm is specified and some potential applications are explored.