ABSTRACT
The previous few chapters have focussed on finite horizon MPC approaches such as GPC and PFC. In both cases, while it is often possible to gain good closed-loop behaviour, the reader may have been somewhat uncomfortable with some of the assumptions made and in particular the theoretical weaknesses associated to these. Of particular importance it is noted that a well-posed optimisation should meet a number of criteria such as the prediction class including the most desired one and the horizons being long enough to capture the whole picture; however, in practice these criteria are only met partially by GPC.
