With demographic change driving the demand for robotic assistance in complex grasping tasks, robot hands are emerging as a promising solution. However, the growing interest in anthropomorphic robot hands is difficult to address: The complex and multifaceted mechanical design creates high entry hurdles for new designers. This slows the growth in the number of experts and complicates finding relevant information, even for experienced designers. This is particularly evident in the conceptual design phase, when determining joint configuration, their coordination, and relative allocation. Guidance in the form of a knowledge-based design assistance system could address these problems by automating the decision-making parts of the design process, presenting solutions with reference to literature and offering recommendations for the conceptual design. This dissertation addresses the question of how such a guidance can be developed. A hypotheses is formulated and three objectives are pursued to test it: (1) develop a literature-linked knowledge base on joint configuration and coordination of tendon-driven rigid-sequential anthropomorphic robot hands to systematically generate conceptual designs, (2) develop a method to identify optimal designs based on user requirements, and (3) develop an automated evaluation method to select the most suitable design from the generated options. For the first objective, a knowledge base is developed, organizing relevant information into a non-redundant list of 23 fields of interest and 30 principal solutions, each linked to its original references. Arranged in morphological boxes, the knowledge base is applied to a methodically derived joint configuration of the human hand to generate approximately 8.5 billion conceptual designs. For the second, a heuristic is developed to identify Pareto-optimal designs based on desired grasp types (performance) and a choice between dexterity and design simplicity (preference). Suitable joint configurations and coordination strategies, drawn from literature, are ranked according to the user’s preference. For the third, the developed control system evaluates designs using a utility analysis of the principal solutions, with scores adjusted by preference and relevance weights to align with the heuristic. The resulting guidance is implemented as a graphical user interface and verified to replicate the intended recommendations. An example application of the guidance presents, to the best of the author’s knowledge, the first task-based design of an anthropomorphic robot hand with references to original sources. By lowering entry barriers in early-stage anthropomorphic robot hand design, this dissertation lays a foundation for accelerating mechanical and control development, with potential to meet growing demand and reduce costs. It also presents the first framework for task-based robot hand design with meaningful results, despite gaps in the state of the art limiting a full analysis.
Daniel Gossen (Wed,) studied this question.