INTELLIGENT CONTROL OF BALL BALANCER AND HELICOPTER SYSTEM

Abstract

The operation of robotic systems to perform complex tasks in dynamic environments has been a crucial area of control. Further, the technological advancements in autonomy, artificial intelligence and robotics have broad applications across society, bringing both opportunities and risks. Most of these opportunities are directly related to path tracking, speed control, maneuverability, and balancing control which are highly affected by the complexity and unpredictable dynamics of the surroundings. Besides, efficient path tracking and balancing control are particularly important for the robots, in order to achieve autonomy without any collision and disturbances. Consequently, the parameters of the mechanical and electronic components need to be monitored and optimized for performing multiple tasks and maintain the reliability of the system. In the view of these aspects, this research identified the combination of intelligent approaches and machine learning methods to achieve unprecedented path tracking and balancing control, continuous monitoring, and robustness by relying solely on onboard computing. The approaches are developed based on multiple control algorithms and are implemented with two-degree freedom operation of ball balancer and helicopter benchmark systems. The controllers are synthesized by first applying the theory of Feedback Linearization and then enhancing their robustness properties. A brief overview of the physical interpretation of the dynamic equations, which is important to the control system designer's understanding of the system, is given. This is followed by further mathematical descriptions of the robust techniques used to augment the basic control law. The research identified the difficulties associated with current control practices and the potential improvements achievable by using nonlinear control. Initially, an intelligent approach for ball balancer position control, and unmanned helicopter trajectory tracking using wavelet fuzzy and evolving type2 quantum fuzzy neural network are developed. The wavelet transform based fuzzy controller overcomes the drawbacks of transparent interpretation of choosing fuzzy rules with techniques available in the literature.

Besides, the evolving type-2 quantum fuzzy neural network is targeted at developing self- organizing and rule growing scenarios with quantum membership functions to overcome the

effects of parametric uncertainties in the benchmark nonlinear system. Further, the development of probabilistic control approaches with randomized algorithms and stochastic

iv

approximations are carried out to estimate the operation of the benchmark systems under random uncertainty conditions. Furthermore, the fault classification based reconfigurable control methods are developed by adapting wavelet transform, machine learning, and intelligent control approaches for both the systems. Finally, the reinforcement learning algorithm-based control approaches are adapted to develop model free controllers with linear quadratic regulator and neural network techniques to perform temporal feedback-based control and interleaved control respectively with the helicopter and ball balancer systems. The

performance of all the developed controllers is validated through simulation studies and real- time analysis for both the ball balancer and helicopter benchmark systems.