Fast NMPC: A reality-steered paradigm: Key properties of fast NMPC algorithms

In this paper, the paradigm of fast Nonlinear Model Predictive Control is recalled. Then a fundamental inequality that conditions the closed-loop stability is derived. Based on this inequality, it is shown that the comparison between different algorithms in the context of Fast NMPC must be based not only on the efficiency per iteration but also on the time needed to perform a single iteration. An illustrative example is used to assess the fact that under some circumstances, it is worth using less efficient algorithms (in conventional sense) that correspond to less amount of computation per iteration and this even when perfect model is used and in a disturbance-free context.