Robust Model Predictive Control for Linear Parameter Varying Systems along with Exploration of its Application in Medical Mobile Robots
| dc.contributor.advisor | Zhang, Chris | |
| dc.contributor.committeeMember | Xu, Peter (Weiliang) | |
| dc.contributor.committeeMember | Singh, Jaswant | |
| dc.contributor.committeeMember | Liang, Xiaodong | |
| dc.contributor.committeeMember | Wu, FangXiang | |
| dc.contributor.committeeMember | Bui, Francis | |
| dc.creator | Hadian, Mohsen | |
| dc.date.accessioned | 2023-04-11T14:25:09Z | |
| dc.date.available | 2023-04-11T14:25:09Z | |
| dc.date.copyright | 2023 | |
| dc.date.created | 2023-03 | |
| dc.date.issued | 2023-04-11 | |
| dc.date.submitted | March 2023 | |
| dc.date.updated | 2023-04-11T14:25:09Z | |
| dc.description.abstract | This thesis seeks to develop a robust model predictive controller (MPC) for Linear Parameter Varying (LPV) systems. LPV models based on input-output display are employed. We aim to improve robust MPC methods for LPV systems with an input-output display. This improvement will be examined from two perspectives. First, the system must be stable in conditions of uncertainty (in signal scheduling or due to disturbance) and perform well in both tracking and regulation problems. Secondly, the proposed method should be practical, i.e., it should have a reasonable computational load and not be conservative. Firstly, an interpolation approach is utilized to minimize the conservativeness of the MPC. The controller is calculated as a linear combination of a set of offline predefined control laws. The coefficients of these offline controllers are derived from a real-time optimization problem. The control gains are determined to ensure stability and increase the terminal set. Secondly, in order to test the system's robustness to external disturbances, a free control move was added to the control law. Also, a Recurrent Neural Network (RNN) algorithm is applied for online optimization, showing that this optimization method has better speed and accuracy than traditional algorithms. The proposed controller was compared with two methods (robust MPC and MPC with LPV model based on input-output) in reference tracking and disturbance rejection scenarios. It was shown that the proposed method works well in both parts. However, two other methods could not deal with the disturbance. Thirdly, a support vector machine was introduced to identify the input-output LPV model to estimate the output. The estimated model was compared with the actual nonlinear system outputs, and the identification was shown to be effective. As a consequence, the controller can accurately follow the reference. Finally, an interpolation-based MPC with free control moves is implemented for a wheeled mobile robot in a hospital setting, where an RNN solves the online optimization problem. The controller was compared with a robust MPC and MPC-LPV in reference tracking, disturbance rejection, online computational load, and region of attraction. The results indicate that our proposed method surpasses and can navigate quickly and reliably while avoiding obstacles. | |
| dc.format.mimetype | application/pdf | |
| dc.identifier.uri | https://hdl.handle.net/10388/14558 | |
| dc.language.iso | en | |
| dc.subject | Linear parameter varying | |
| dc.subject | Model predictive control | |
| dc.subject | Robust control | |
| dc.subject | Recurrent neural networks, Robots, Support vector machine | |
| dc.title | Robust Model Predictive Control for Linear Parameter Varying Systems along with Exploration of its Application in Medical Mobile Robots | |
| dc.type | Thesis | |
| dc.type.material | text | |
| thesis.degree.department | Biomedical Engineering | |
| thesis.degree.discipline | Biomedical Engineering | |
| thesis.degree.grantor | University of Saskatchewan | |
| thesis.degree.level | Doctoral | |
| thesis.degree.name | Doctor of Philosophy (Ph.D.) |