Water-level reference planning for automated irrigation channels via robust MPC

In a water-level reference planning problem for automated irrigation channels, the object is to steer the system along trajectories that satisfy hard constraints on the regulated water levels and the flow commands produced by the underlying feedback controllers. The approach proposed here involves nominal trajectories that are devised off-line accounting for a forecast (i.e. schedule) of the demand (i.e. load input). The load actually drawn during operation is inevitably a perturbed version of this schedule. To achieve a level of robustness, insomuch as the deviation from nominal trajectories respects the constraints in the presence of such uncertainty, a Model Predictive Control (MPC) scheme is developed to adjust the nominal reference plan on-line. This receding horizon scheme involves time-varying constraints on the error relative to the nominal trajectories and a novel on-line approach to constraint tightening, which yields a robust feasibility property. Simulation results for a stretch of an automated irrigation channel are discussed.