A COMPARATIVE STUDY ON EMBEDDED MPC FOR INDUSTRIAL PROCESSES
Adriano Silva Martins Brandão1; Daniel Martins Lima1; Marcus Vinicius Americano da Costa Filho2; Julio Elias Normey-Rico1
1 Universidade Federal de Santa Catarina; 2 Universidade Federal da Bahia
Baixar PDF doi:10.20906/CPS/CBA2018-0569
Resumo
Predictive control algorithms have great acceptance in the process industry because of its ability to handle multivariable and constrained systems. The presence of constraints greatly increases the computational cost of solving these problems, which has motivated many works in recent years to develop automatic code generation tools for MPC applications. These developments enable the execution of advanced controllers in low-cost embedded systems, which are increasingly applicable to industrial processes due to advances in the development of Internet of Things (IoT) devices. Although several libraries that implement embedded MPC exist, there are not many comparative studies, which makes it difficult to choose among them. The present work aims to compare three of these tools (CVXGEN, FalcOpt and muAO-MPC), to list advantages and disadvantages of some of these tools, to test the performance on an embedded system and to present suggestions of uses and restrictions for applications.
Palavras-chave: Model Predictive Control; Embedded hardware; Code-generation