Coutts, Aaron (2021) Optimal fuel cost controller design for a helicopter/twin rotor system. [USQ Project]
|
Text (Project)
COUTTS Aaron dissertation_redacted.pdf Download (3MB) | Preview |
Abstract
The TRMS (Twin Rotor Multi-Input-Multi-Output [MIMO] System) 33-949 system is a small-scale model of a helicopter, exhibiting many similar characteristics, except for limited degrees of freedom. The non-linearities, as well as cross-couplings between the inputs and outputs of such a system make the model a useful basis for designing, testing, and deploying a broad range of control algorithm implementations, and allow for experimentation and exploration of various optimisation techniques.
Many Machine and Deep Learning techniques have been discovered that allow researchers to optimise control algorithms based on multiple objectives. Research was conducted to find a method of optimisation that allowed the search of a large space of candidate solutions to minimise an objective function of error and cost. The evolutionary-based genetic algorithm was suitable, demonstrating good results on the TRMS system. Memetic Algorithms (MA) were discovered to have been applied in system identification, but not yet applied in finding optimal control parameters for a TRMS system. A MA is of the same family of algorithms as a Genetic Algorithm (GA), so the performance of the memetic algorithm is directly comparable to the established research on GA and its application to the TRMS.
Implementation and testing has revealed that the MA does outperform an equivalent GA (the same algorithm, with the Local Search removed) in every test case, with typical fitness value improvements of anywhere from 5% to 120% (and greater improvements are likely as observed from the results data). Of particular interest was the capability of MA for finding simultaneous decoupling and control solutions. Test results have indicated however, that significant cross-couplings still exist, particularly in the pitch-to-yaw path. Therefore, the MA has been useful for designing a cost/error optimal controller, especially for Single-Input-Single-Output (SISO) applications, but as with many developments in AI it certainly is not without limitations.
|
Statistics for this ePrint Item |
| Item Type: | USQ Project |
|---|---|
| Item Status: | Live Archive |
| Faculty/School / Institute/Centre: | Historic - Faculty of Health, Engineering and Sciences - School of Mechanical and Electrical Engineering (1 Jul 2013 - 31 Dec 2021) |
| Supervisors: | Wen, Paul; Song, Bo |
| Qualification: | Bachelor of Engineering (Electrical and Electronic) |
| Date Deposited: | 03 Jan 2023 00:51 |
| Last Modified: | 26 Jun 2023 01:05 |
| Uncontrolled Keywords: | TRMS, helicopter, machine learning, deep learning, algorithm, memetic algorithm, genetic algorithm, decoupling, control |
| URI: | https://sear.unisq.edu.au/id/eprint/51807 |
Actions (login required)
 |
Archive Repository Staff Only |
