Reference Governors: On-Line Set-Point Optimization Techniques for Constraint Fulfillment

Abstract

This dissertation presents a control technique to cope with set-point tracking problems in the presence of input and/or state constraints. The main idea consists of feeding to a conventional controller artificial set-points, which are calculated in real-time by manipulating the desired reference trajectory. For this reason, the resulting control tool is called {em reference governor} (RG). Set-point manipulation is performed on-line through an optimization procedure. This attempts at minimizing a performance index, which is evaluated by predicting the future evolution of the system. The RG is a high-level intelligent control module which supervises conventional controller operation, by ``smoothing out'' the reference trajectory when abrupt set-point changes would lead to constraint violation. The proposed control scheme is computationally light and easily implementable on low-cost hardware, and is general enough to cope systematically with different constrained tracking problems. We develop here theory and present simulation results of reference governors for linear, nonlinear, uncertain, robotic, and teleoperated control systems.