This paper proposes a hierarchical decentralized reference governor for constrained cascade systems. The reference governor (RG) approach is reformulated in terms of receding horizon strategy such that a locally receding horizon optimization is obtained for each subsystem with a pre-established prediction horizon. The algorithm guarantees that not only the nominal overall closed-loop system without any constraint is recoverable but also the state and control constraints are satisfied in transient conditions. Also, considering unfeasible reference signals, the output of any subsystem goes locally to the nearest feasible value. The proposed dynamic constraint tightening strategy uses a receding horizon to reduce the conservatism of conventional robust RGs. Moreover, a decentralized implementation of the algorithms used to compute tightened constraints and output admissible sets is introduced that allow to deal with large scale systems. Furthermore, a set of dynamic constraints are presented to preserve recursive feasibility of distributed optimization problem. Feasibility, stability, convergence, and robust constraint satisfaction of the proposed algorithm are also demonstrated. The proposed approach is verified by simulating a system composed of three cascade jacketed continuous stirred tank reactors.
