Ball and Plate MPC Control of a 6 DOF Stewart Platform

MPC is a cutting-edge control technique which is nowadays deeply studied, since it permits to overcome several disadvantages related to simpler and more traditional control methods. In industry, Proportional Integral Derivative controllers are still widespread, due to their low complexity and tuning procedure. However, they do not have the capability to cope with coupled multi-input multi-output systems, as well as constrained and non-linear ones. On the other hand, Model Predictive Control is able to deal with these features at the price of computational complexity. The latter is related to its structure: MPC is indeed an optimization-based control technique which require to solve an optimization problem within every timeslice, and this implies high computational burden. Therefore, its use has been so far mainly limited to academic applications or chemical processes, where the timescale is quite large (around the minute). However, recent advances in both embedded hardware and MPC software libraries have made Model Predictive Control a viable candidate also for systems with fast dynamics, e.g. robotics. With regard to this, the objective of this master thesis is to explore the suitability of Model Predictive Control for motion control applied to fast-timescale systems. The hardware under test consists in a Stewart platform, i.e., a robotic system governed by six servo-actuators which can move a flat plate. The implemented controller aims at balancing a ball towards the middle of the plate. \\ The used MPC library is based on acados, a collection of solvers for fast embedded optimization. Although it is implemented in C, acados is able to interface with higher-level interfaces, such as Matlab, Python and C++, feature which makes acados really flexible. On the other hand, this library provides also efficiency, as its core is written in a low-level language. \\ The project firstly develops simulations by the use of Matlab, in order to observe the feasibility of MPC in terms of computational time and capability to control the system. In a second time, the hardware is employed to verify the simulation results. This includes the setting of the communication in accordance to the serial protocol developed for the embedded systems platform within the hardware.