J. M. Maestre

Distributed Model Predictive Control: An Overview and Roadmap of Future Research Opportunities

Model-predictive control (MPC) is an optimization-based control technique that uses 1) a mathematical model of a system to predict the systems behavior over a given horizon, 2) an objective function that represents what system behavior is desirable, 3) a mathematical formalization of operational constraints that have to be satisfied, 4) measurements of the state of the system at each time step, and 5) any information regarding upcoming disturbances that may be available. This article surveyed and categorized 35 distributed MPC approaches. Subsequently, several of the insights gained from the survey were presented. This study provides a picture of what features have received more or less attention over the last years, bringing about potential research niches for new approaches.

Distributed Model Predictive Control Made Easy

The rapid evolution of computer science, communication, and information technology has enabled the application of control techniques to systems beyond the possibilities of control theory just a decade ago. Critical infrastructures such as electricity, water, trafficand intermodal transport networks are now in the scope of control engineers. The sheer size of such large-scale systems requires the adoption of advanced distributed control approaches. Distributed model predictive control (MPC) is one of the promising control methodologies for control of such systems. This book provides a state-of-the-art overview of distributed MPC approaches, while at the same time making clear directions of research that deserve more attention. The core and rationale of 35 approaches are carefully explained. Moreover, detailed step-by-step algorithmic descriptions of each approach are provided. These features make the book a comprehensive guide both for those seeking an introduction to distributed MPC as well as for those who want to gain a deeper insight in the wide range of distributed MPC techniques available.

Integration of service robots in the smart home by means of UPnP: A surveillance robot case study

Despite the large number of electronic devices that coexist in homes, only a few of them can be easily integrated in the same network. Given that smart home applications are based on the integration of many heterogeneous devices in the same network, the lack of interoperability can become a major issue in the development of advanced services. In this paper we tackle this issue developing an adapter to integrate the service robot Rovio into a smart home by means of Universal Plug and Play (UPnP). Different advanced services are designed to explore the possibilities derived from the integration of service robots in the smart home.

A distributed MPC scheme with low communication requirements

In this work, we consider the problem of controlling two linear systems coupled through the inputs. We propose a novel distributed model predictive control method based on game theory in which two different agents communicate in order to find a cooperative solution to the centralized control problem. We assume that each agent only has partial information of the model and the state of the system. The class of systems considered arises naturally in multi-input multi-output processes in which a transfer function model is obtained using standard identification techniques. The performance and the robustness of the proposed control scheme with respect to data losses in the communications are illustrated by extensive simulations.

Distributed MPC: a supply chain case study

This work focuses on the application of distributed model predictive control to find the optimal decision variables to maximize profit in supply chains. A reduced version of the MIT beer game made of only two elements is taken as an application example. Three controllers, i.e., a standard centralized model predictive controller, an distributed non-cooperative model predictive controller and a recently proposed distributed scheme based on a cooperative game are applied to maximize profit. The properties of these controllers are compared extensively under different simulation scenarios.

Distributed receding horizon Kalman filter

In this paper a distributed version of the Kalman filter is proposed. In particular, the estimation problem is reduced to the optimization of a cost function that depends on the system dynamics and the latest output measurements and state estimates which is distributed among the local subsystems by means of dual decomposition. The techniques presented in the paper are applied to estimate the position of mobile agents.

Application of an In-Line Storage Strategy to Improve the Operational Performance of Main Irrigation Canals Using Model Predictive Control

AbstractStoring water in main irrigation canal reaches could be an influential strategy to improve the existing operational activities in the irrigation canals. However, the control of such a canal system will become much more complicated due to freeboards of the reaches temporarily decreasing. In this paper, Model Predictive Control (MPC) is applied to control the water level of an accurate model of a realistic main canal, which consists of 13 canal reaches, using an in-line storage operational strategy. Four different test scenarios are selected to cover a range of conventional to unconventional operational strategies by imposing limitations on the head-gate opening. Different target bands are created between the predefined allowed maximum and minimum water level for the canal reaches and the MPC is obliged to keep the water levels within these ranges. The results show that the in-line storage improves current operational performance of the canal system by compensating the existing delay times of flow t...

An Iterative Design Method for Coalitional Control Networks with Constraints on the Shapley Value

Abstract In this work, we introduce a new iterative design method for a coalitional control scheme for linear systems recently proposed. In this scheme, the links in the network infrastructure are enabled or disabled depending on their contribution to the overall system performance. As a consequence, the local controllers are divided dynamically into sets or coalitions that cooperate in order to attain their control tasks. The new design method allows the control system designer to include new constraints regarding the game theoretical tools of the control architecture, while optimizing the matrices that define the controller.

Distributed MPC based on a cooperative game

In this work, we focus on the problem of stabilization of two constrained linear systems coupled through the inputs by two different agents which communicate in order to take a decision assuming that each agent only has partial information of the model and the state of the system. We extend previous results on distributed model predictive control and provide sufficient conditions that guarantee practical stability of the closed-loop system as well as an optimization based procedure to design the controller so that these conditions are satisfied. The theoretical results and the design procedure are illustrated through a simulation example.

Stock management in hospital pharmacy using chance-constrained model predictive control

One of the most important problems in the pharmacy department of a hospital is stock management. The clinical need for drugs must be satisfied with limited work labor while minimizing the use of economic resources. The complexity of the problem resides in the random nature of the drug demand and the multiple constraints that must be taken into account in every decision. In this article, chance-constrained model predictive control is proposed to deal with this problem. The flexibility of model predictive control allows taking into account explicitly the different objectives and constraints involved in the problem while the use of chance constraints provides a trade-off between conservativeness and efficiency. The solution proposed is assessed to study its implementation in two Spanish hospitals.