Robust MPC for disturbed nonlinear discrete-time systems via a composite self-triggered scheme

In this paper, an aperiodic formulation of model predictive control (MPC) with composite self-triggered mechanism is proposed to reduce communication and computational load, while retaining a desired control performance. Concretely, a less conservative error estimation between the actual and predicted states is utilized to design the triggering mechanism, leading to a sufficient reduction in the frequency of solving optimal control problems. By introducing a contraction mapping function, the prediction horizon shrinking strategy is proposed to shorten the length of prediction horizon as the state close to the target set, which further reduces the computational complexity of optimal control problems at each triggered instant. Two involved tuning parameters, performance factor and horizon shrinking factor, can be used to achieve a sensible compromise between system properties and energy saving. The sufficient conditions to guarantee the validity and implementations of the proposed algorithm are indispensable and hence investigated technically with reference significance for nonlinear MPCs. The proposed algorithm is shown to ensure recursive feasibility and robust stability, and its efficiency is illustrated through a numerical example.