Explicit Model Predictive Control of an Automated Storage / Retrieval System

This thesis is written with the objective to research the possibilities of explicit Model PredictiveControl applied to a practical implementation. In particular, in this work the realistic example of anAutomated Storage / Retrieval System has been considered and a control-oriented model has beenderived. This model has allowed the design and implementation of an Explicit Model PredictiveControl law, from which the advantages and disadvantages can be researched. The project hasbeen developed in partnership with the Politecnico di Milano and the Consiglio Nazionale delleRicerche.The rst step of the project focused on the analysis of a generalized description of the AutomatedStorage / Retrieval System and on the study of the physical equations describing its behavior. Inparticular, in view of the design specications, the dynamics of the system have been identied andthe manipulated and controlled variables have been dened. Subsequently, the physical equationsdescribing the system have been analyzed, and a non-linear hybrid model has been obtained.After the derivation of the model, dedicated software has been used to explore possibilities toconvert the non-linear hybrid model into a linear and easily optimizable form. Subsequently, thesteps to compute an explicit control law using these linear hybrid models have been explained.Additionally, a custom-made program has been written to compute the explicit Model PredictiveControl law for Automated Storage / Retrieval Systems based on the non-linear system.The generalized form has been used in two test cases from which the benets and the advantagesof applying an explicit Model Predictive Control in industrial systems are concluded. The mainndings are that the explicit law is feasible for systems with a limited amount of inputs and states,due to the fact that the online and oine computation of the control law in the explicit formcorrelate signicantly with these parameters.