Design of multiple-functions controller based on machine learning

• Main Author: Syafiq Fauzi, Kamarulzaman
• Format: Thesis
• Language: English
• Published: 2015
• Subjects: T Technology (General); TJ Mechanical engineering and machinery
• Online Access: http://umpir.ump.edu.my/id/eprint/13559/1/Design%20of%20multiple-functions%20controller%20based%20on%20machine%20learning.wm.pdf; http://umpir.ump.edu.my/id/eprint/13559/

Human has the ability to learn and decide their action based on experiences when confronting a problem. Human decision often involves multi-functionality, where multiple control functions are applied for achieving a single goal. Conventional control often involves human in providing commands which mostly depends on the human decision. However, these decisions commonly involve single control function where multi-functionality can not be provided without human assistance. Learning Control helps a machine constructs its own control knowledge autonomously through operation experiences. The development of Control Knowledge through Learning Control would require a period of training that could involve a number of failures among successful attempts. The Control Knowledge obtained is usually limited to single control function based on the training environment with less flexibility in varying environment. Learning Control Systems with multiple functions could provide a wider range of control options against any environment. In this research, Learning Control System with multi-functionality is designed and developed. Here, application of Learning Control with multi-functionality provides a more human-like control operation with ability to adapt and consider the surrounding environment during control operation. The designs were evaluated through experiments and simulations where results confirm the effectiveness of the designed system. Through these results, the designs of multi-functions Learning Control may provide a safer and reliable control on control devices including complex non-linear control device.