Control configuration selection for linear stochastic systems

• We introduced an interaction measure for stochastic systems for the first time.
• We explained how to use the proposed interaction measure for control structure selection.
• The proposed control configuration selection method can be used to propose richer control structures.
• A method for sensor and actuator placement is proposed for stochastic systems.

An appropriate control configuration selection is identified as one of the key prerequisites for attaining the control objectives in industrial practices. To select a suitable control configuration, it is important to determine which variables should be measured and how the process should be actuated. Therefore, the first step is to determine the optimal locations for the sensors and actuators. For the multivariable processes, this step is followed by choosing the appropriate input and output pairs for the design of SISO (or block) controllers. This is due to the popularity of the distributed and decentralized control in industrial control systems. These issues, which have been studied extensively for deterministic systems, have not been closely studied for stochastic systems. In this paper however the problem of control configuration selection is studied for the linear stochastic systems. The problem of selecting the sensor locations for stochastic systems is viewed as the problem of maximizing the output energy generated by a given state and for the actuator locations is viewed as the problem of minimizing the input energy required to reach a given state. Furthermore, a gramian-based interaction measure for control structure selection of multivariable stochastic systems is proposed. This interaction measure can be used to propose a richer (sparse or block diagonal) controller structure for distributed and partially decentralized control.