Model predictive control for nonlinear systems modelled with neutral networks
Model predictive control (MPC) is a popular and an advance control technique for linear system with hard constraints. However, in the case of non-linear system, using MPC to control may pose difficulties. This is because MPC may not be able to handle the non-linear system’s complexity. With new tech...
Saved in:
Main Author: | |
---|---|
Other Authors: | |
Format: | Final Year Project |
Language: | English |
Published: |
2017
|
Subjects: | |
Online Access: | http://hdl.handle.net/10356/71419 |
Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
Institution: | Nanyang Technological University |
Language: | English |
Summary: | Model predictive control (MPC) is a popular and an advance control technique for linear system with hard constraints. However, in the case of non-linear system, using MPC to control may pose difficulties. This is because MPC may not be able to handle the non-linear system’s complexity. With new technologies and ideas emerging, the interest in MPC arises as it has potential to conquer this weakness and exhibits its unique strength. Neural Network (NN) is a type of classifier to predict its output and can be used as process model for many control problems. In addition, the efficient of a non-linear MPC is highly related to the properties of NN.
In this research, different neural network techniques were compared through cross validation. Two layer NN using back-propagation as its learning algorithm yield the lowest errors. The number of nodes in two layer NN was evaluated through grid search and learning rate was investigated. In MPC, WMR is used as the problem description and it was observed that the MPC has a good trajectory tracking performance while handling system with nonholonomic constraints. |
---|