Data-Driven Model Predictive Control of Cable-Driven Parallel Robots

The main objective of this thesis is to implement a data-driven methodology for MPC in CDPRs (Cable Driven Parallel Robots). The increasing availability of data and the scalability with which they are stored can be used in such different fields. In particular both with the technological development and the advent of extremely complex artificial intelligence algorithms, it becomes fundamental for engineering aims. Data driven control is precisely based on this huge availability of data, and it consists of directly using input and output data from the system to design the controller. In this thesis the developed Data Driven MPC consists on learning from data to obtain the space-state system model by using system identification. After that an MPC is applied on the identified model to control the end effector position. As mentioned before the system analyzed is a Cable Driven Parallel Robot (CDPR) with a massive end-effector. The aim is to control the mass position through three cables, and hence three actuators. As the result CDPR's system identification has been successfully achieved by using MATLAB's System Identification Toolbox, and then the MPC is applied as the offline control technique. Another engineering challenging goal is the Online System Identification, i.e. a System Identification Technique developed by using the updating of input data and their respective updating of output data. This metodology is like an online fitting curve mechanism, which is based on the Recursive Least Square (RLS) Algorithm. This thesis proposes a sort of Online System Identification by using the RLS algorithm, but it has to be improve and tested more.