On the design of reconfigurable two layer hierarchical control systems with MPC

Abstract

This paper deals with the design of a hierarchical two-layer controller. At the high level, a robust MPC regulator works at a slow frequency rate and computes the ideal control signals needed to suitably control the plant. At the low level, a number of already controlled actuators drive the plant at a faster time rate and track the desired control signals as accurately as they can, in accordance to their dynamics. The discrepancy between the ideal control actions computed at the high level and those actually provided to the plant gives rise to a robustness issue for the overall control system. To tackle this problem, we propose to model such a discrepancy as a disturbance term to be rejected in the design of the high level controller by means of the small-gain approach. Moreover, setting an optimization procedure at each time step, the proposed algorithm also allows for an easy reconfiguration of the control system in the occurrence of an actuator addition/replacement.

Tracking control of Wiener models with hierarchical and switching model predictive control

“…In [14], a hierarchical controller has been proposed, where at each sampling instant, the upper level unit can select a set of active actuators forming the lower level controller so as to meet secondary objectives, such as control effort distribution and performance requirements. This idea has been extended in [15] to deal with reconfigurability issues. 2 B. PICASSO, C. ROMANI AND R. SCATTOLINI In this paper, stemming from preliminary results developed in [16,17], a two-layer hierarchical control structure is designed for the asymptotic tracking of a constant reference signal in a discretetime, multi-input, multi-output Wiener model, see Figure 1.…”

“…In , a hierarchical controller has been proposed, where at each sampling instant, the upper level unit can select a set of active actuators forming the lower level controller so as to meet secondary objectives, such as control effort distribution and performance requirements. This idea has been extended in to deal with reconfigurability issues.…”