Receding horizon control using modified iterative dynamic programming and neural network models

The basic idea of the contribution is to replace a mathematical model of the process by an equivalent neural network (NN) that mimics the phenomenological model and it is used as a process predictor in the modified iterative dynamic programming control (IDP) algorithm. IDP is a very useful technique for solving unconstrained and constrained dynamic optimisation problems. nle original IDP method is developed for continuous systems within state space formulation. The modified algorithm uses a learned NN as a process predictor. The algorithm modifications resulting from this type of the models include several important issues that arise from the use of discrete-time and input-output model formulations. Moreover, there are also some significant problems that are not to be overlooked stemming from the receding horizon implementation of the method. The contribution discusses all these issues. The benefits of the proposed approach are small number of iterations required to converge to global optimum, ability to handle multivariable constrained systems and significant time reduction compared to the original IDP method.