The main result of our new method is to convert an output constraint into an input constraint for a nonlinear system. Then, standard saturated input design can be used as an anti-windup design for instance which is known to be efficient to enlarge a closed-loop system stability domain in presence of control saturations. We only address the single input single 'saturated' output problem in this paper.
I. INTRODUCTIONOutput and Input hard constraints problems arise in most of the control applications because achieving good performance often requires to reach the inherent physical and safety limitations of a concrete system. Even if one has to cope with possible future destabilizing effects, the application of a saturation placed just after an input is a simple and sure strategy to handle an input constraint whereas there does not exist an equivalent element for an output limitation. That's probably why the input constraint problem has received more attention in the past few decades (see for instance [4]). Existing solutions to these problems can be divided into two groups : those who predict the future of the closed loop trajectory and check if the constraints will be violated and the other, which at each time 'do their best' to avoid the constraints. Concerning Linear systems, not only future prediction is easier but we can also apply some dedicated LMI-based methods [6], [2]. Moreover, state and input constraints problems are very close since it is possible to apply a relationship between a constrained output and an induced constrained input (whose constraints are state dependent) when the system is perfectly known and discretized [3], [1]. In this paper, we propose to generalize this idea to nonlinear systems by transforming an output constrained nonlinear system into an input constrained nonlinear system whose constraints are state dependent. As we will see our method can still be applied when some disturbances are applied to the nonlinear system and does not require to discretize the system but contrary to the linear case it can be slightly conservative. By comparison to existing results based on nonlinear model predictive control [5], [9], [7], [10] (to cite a few), our method does not use prediction and can thus be easily implemented ; however, it is fair to say that we do not address several output constraints problems (where the number of inputs is inferior to the number of constrained outputs) in this preliminary work. Other possible and very useful methods [8], [11], [12] address output constraints problems for nonlinear systems of special form and/or try to escape the constraints while