Generation, evaluation and dimensional synthesis of planar hybrid manipulators: a unified approach

Abstract

Tremendously increasing variations in the robotic applications has led to the importance of utilizing hybrid configurations for the design of robotic manipulators. The motivation lies in the fact that serially connected links are normally selected for larger manipulability, and parallel manipulators are utilized for better positional accuracy. Common choices of hybrid configurations include either serially connected parallelograms or a loop inserted in a serial manipulator at different locations. With various hybrid combinations possible, even with a fixed number of degrees-of-freedom (dof), the experience and previous knowledge play key roles in the selection of a basic configuration. Once a basic structure is selected, it is followed by some systematic/optimal geometric synthesis, kinematic and/or dynamic design methodologies. However, selection of a basic structure of a manipulator configuration itself is critical enough, which needs further attention. Literature digests are presented for quick visualization of the research directions related to the proposed work, arranged with respect to their era and area. To provide solutions for the selection of a hybrid morphology, performance analysis platform and optimal dimensional synthesis of planar hybrid manipulators, the work has been planned in four phases — random morphology generation, retrieval of kinematic information, unified performance evaluation criteria and evolutionary synthesis for planar hybrid manipulators. ‘Morphology’ here is referred to a basic kinematic framework for a manipulator. A large set of hybrid configurations are randomly generated for a range of number of degree-of-freedom and evaluated through a proposed unified approach for kinematic performance analysis. This provides a most suitable planar hybrid morphology selection for a given set of task-space locations. In the first phase, the work is to randomly generate the variations in the possible morphologies of planar hybrid manipulators with any number of links. For this, a stipulated range of number of degrees-of-freedom had been considered and all the possible variations in manipulators with links connected in series and/or loops had been explored with the help of introduction of Mechanism Assembly Matrix (MAM). Each element of the matrix signifies a joint between the row-indexed link and the column-indexed link. After generation of MAM, the major task was to retrieve the linkage information corresponding to each generated morphology. This information was required to perform the analysis in next phase. Major challenge in retrieval of such information is faced due to variations in the randomly generated mechanisms. For this, a graph-theorybased algorithmic framework had been developed for the determination of kinematic information. This includes all serially connected links, link numbers, number of loops, loops locations and sizes, base location and the end-effector link of the generated hybrid manipulators. The development of general performance evaluation criteria has gone through various stages, to develop a unified approach for computation of randomly generated morphologies. Due to the variations in the basic configurations, one of the major challenge was the development of kinematic models corresponding to each generated linkage. A unified approach is proposed for kinematic modelling and performance evaluation criteria of generated linkages. The proposed approach is verified for specific cases of planar hybrid morphologies. Comparative study is detailed based upon the computational efforts, extendibility and modularity of the approach. Final phase of the work deals with the integration of methodologies proposed for three phases, and utilizing the resulting overall approach for evolutionary design of all selected configurations. The evolutionary algorithm had been simulated for selection of the best possible manipulator morphology for a specified task. Jacobian based conditioning index has been considered as the performance criteria for optimal dimensional synthesis of all generated planar hybrid manipulators. The overall scheme of generation and selection of basic planar hybrid manipulators has been applied to the realistic applications of arm-rehabilitation and sit-to-stand assistive mechanism synthesis.