Title:Reference tracking stochastic model predictive control over unreliable channels and bounded control actions

Abstract:A stochastic model predictive control framework over unreliable Bernoulli communication channels, in the presence of unbounded process noise and under bounded control inputs, is presented for tracking a reference signal. The data losses in the control channel are compensated by a carefully designed transmission protocol, and that of the sensor channel by a dropout compensator. Our proposed class of feedback policies is parametrized in terms of the disturbances appearing in the dynamics of the dropout compensator. A reference governor is employed to generate a trackable reference trajectory and mean-square boundedness of the error between the states of the system and that of the reference governor is achieved with help of stability constraints. The overall approach yields a computationally tractable quadratic program, which is solved iteratively online.