Article révisé par les pairs
Résumé : This paper gives a general way to take into account, by an appropriate design of the controller, any discrepancy which can occur between the actual inputs of a process and the desired outputs of its controller. This yields to the so-called conditioned control algorithms. The conditioning technique is described for multiple-input-multiple-output nonlinear controllers. Application to complex control structures is explained. The notion of a self-conditioned controller is defined in order to simplify the implementation of the conditioning technique. Some considerations about the stability of conditioned systems are given. The automatic initialization of any control algorithm is given as an application of the method. It shows that bumpless transfer can be achieved.