Learning-Based Methods for Decision-Making, Planning & Control

Accelerating manufacturing speeds and capabilities remain at the forefront of aerospace research. Aerostructures are currently assembled using large static jigs composed of fixturing elements, leading to slow production cycles due to the lack of reconfigurability. As the industry begins to embrace robotic solutions due to the rise of their capabilities in manufacturing tasks, the question has arisen as to how to develop intelligent and autonomous behaviour in these robotic fixtures. In this thesis, we harness the capabilities of these robotic fixtures to build autonomous systems that can perform three of the major operations that are associated with the development of autonomous jigs. In partnership with Airbus, these areas have been identified as the ones that stand the most to benefit from autonomous systems and the ones that would have the largest impact on the manufacturing process. Firstly, we look to develop an algorithmic approach for determining the locations for the fixturing elements to constrain a component during manufacturing tasks. Secondly, we tackle the problem of constrained manipulation of components, an operation that is frequently used across different tasks in manufacturing and is particular importance to ensure that no damage happens to the component during operation. Finally, we examine the problem of autonomous robot assembly of components, in particular looking at the use of deterministic assembly for aerospace components. This thesis draws together a variety of fundamental concepts and formal methods in robotics in partnership with machine learning techniques to enhance the capabilities of the proposed solutions. With guidance from Airbus and applications rooted in manufacturing, this thesis presents novel approaches to tasks within autonomous jigs to stimulate the discussion regarding autonomous rigging and show that they belong within the factories of the future.