Embedded model predictive control on a microcontroller.
Su, Yong Yao.
Date of Issue2008
School of Electrical and Electronic Engineering
Model Predictive Control (MPC) has become an established control technology due to its powerful ability in constraints handling. The ability to solve MPC problems online becomes critical for application that requires fast response time especially embedded system that has limited computational resource. The key component of model predictive control is the solving of quadratic programming problem (QPP). Interior point method (IPM) and active set method (ASM) appear to be the most efficient approaches for solving general QPPs. This project compares the performances of the methods on 32-bit microcontroller in 4 aspects, i.e. computational complexity, storage, convergence speed, and numerical error. The findings show that, in general, ASM gives lower complexity and shorter computation time. However, it uses larger memory space and does not produce converged results for some of the QPPs that have no feasible point at (0,0,…,0). On the other hand, IPM uses less memory space and able to produce converged results for all the QPPs that have feasible point at (0,0,…,0). In addition, formulation of a basic MPC problem for Cessna Citation 500 aircraft is discussed and it is used to verify the algorithm of the methods that are implemented on the STM32 microcontroller. Lastly, the peripherals’ configuration of the STM32 microcontroller has been set up and ready for implementation of a MPC controller to balance the pendulum of apparatus PP-300.
DRNTU::Engineering::Electrical and electronic engineering::Control and instrumentation::Control engineering
Final Year Project (FYP)