Inferential networked H∞ control with accessibility constraints in both the sensor and actuator channels

The predictor and controller design for an inferential control scheme over a network is addressed. A linear plant with disturbances and measurement noise is assumed to be controlled by a controller that communicates with the sensors and the actuators through a constrained network. An algorithm is proposed such that the scarce available outputs are used to make a prediction of the system evolution with an observer that takes into account the amount of lost data between successful measurements transmissions. The state prediction is then used to calculate the control actions sent to the actuator. The possibility of control action drop due to network constraints is taken into account. This networked control scheme is analysed and both the predictor and controller designs are addressed taking into account the disturbances, the measurement noise, the scarce availability of output samples and the scarce capability of control actions update. The time-varying sampling periods that result for the process inputs and outputs due to network constraints have been determined as a function of the probability of successful transmission on a specified time with a Bernoulli distribution. For both designs H-infinity performance has been established and linear matrix inequality (LMI) design techniques have been used to achieve a numerical solution.