Reference Governors

CU Engineering Building

Comic-style illustration of the principle behind Reference Governors

Original artwork by , commissioned by Marco M. Nicotra and Emanuele Garone.

Reference Governors are a family of add-on control units that act as an intermediary layer between the low-level control law and the high-level guidance system. Their objective is to generate a reference signal that is as close as possible to the one requested by the guidance layer, while simultaneously ensuring that the transient response of the closed-loop system is safe and does not violate constraints.

The introduction of a Reference Governor significantly simplifies the design of both the low-level control law, which can focus on stabilizing the underlying system, and the guidance layer, which can define the desired outcome in fairly general terms. This is achieved by modulating the reference of the closed-loop system so that the transient behavior is always compatible with the system constraints.

Due to their specific formulation and mathematical properties, Reference Governors tend to be simple to interpret, systematic to use, and computationally inexpensive to run. This makes them a prime candidate for augmenting existing solutions that were not specifically designed to handle constraints.

Real-Time Iteration Governors

The Real-Time Iteration Governor (RTIG) is a innovative tool that is used to manipulate the reference of a Real-Time Model Predictive Control law to ensure that the solution estimate error does not exceed a given threshold. This approach paves the way for computationally inexpensive MPC strategies that can implement an optimal control sequence, while having guaranteed numerical convergence properties due to the RTIG.

Collaborators: 1
1. University of Michigan

Students: Terrence Skibik

Funding:  (Award Number: 1904441)

Explicit Reference Governors

The Explicit Reference Governor (ERG) is a continuous-time implementation strategy that provides the same benefits as the more traditional discrete-time schemes without having to actually solve an online optimization problem. This is done by manipulating the derivative of the applied reference and taking advantage of invariance principles.

Collaborators: 1

1. Université Libre de Bruxelles

 

Showcase