Workplace analysis for mobile manipulation

Abstract

The dissertation has developed a general methodology for workspace analysis of mobile manipulation. The analysis of workspace for mobile manipulators is crucial in designing their structures and planning their motions for specific tasks. Traditional numerical solutions with simple algorithms often suffer from high computational costs, low accuracy, and limited portability. Analytical solutions, while more effective, struggle to handle the complex non-linear properties of mobile manipulators' workspace distribution, especially with high degrees of freedom and infinite extension in a 6-dimensional task space. Hence, there is a pressing need for a novel theoretical research approach to workspace analysis that accounts for generic structure parameters. To address this challenge, a mathematical model capable of describing the 6-dimensional workspace distribution is proposed. This model provides a quantitative representation of property distribution and facilitates motion planning for various mobile operations. Unlike redundant stationary manipulators with fixed bases, mobile manipulators equipped with wheel joints exhibit unique kinematic contributions to end-effector motion and infinite extension of workspace. This behavior can be described as an infinite high-dimensional manifold in the task space influenced by multiple degrees of freedom and other constraints. To capture the inherent complexity, we propose a singularity analysis based on interval estimation theory, coupled with a workspace model utilizing deep neural networks, which replaces the conventional discretization representation known as the "binary map." The proposed workspace model can represent the intricate high-dimensional manifold and can be used for practical purposes such as motion reachability verification and motion planning. Additionally, the workspace model can identify regular patterns in specific property distributions and extrapolate them to the infinite task space, accommodating diverse structure parameters of mobile manipulators. Concrete examples of motion planning tasks, such as mobile painting and surface disinfection, are demonstrated to validate the effectiveness of the application of mobile manipulation. To verify the proposed methodology, multi-goal constraint optimization problems are formulated, and an automated tool planning software system is developed based on CAD models of surfaces, tool deposition models, optimization criteria, and other constraints. The dissertation presents theoretical analysis for pose and orient reachability, using the proposed workspace model of mobile manipulators. It includes detailed illustrations and analyses of workspace types and properties, such as reachable space and singularities. The effectiveness of the workspace model is demonstrated, and methodologies for analyzing properties are presented. Motion planning cases for mobile manipulations is performed through simulation and experimental tests. Finally, the results are analyzed and evaluated, showing satisfactory outcomes that validate the effectiveness of the proposed methodology.