| Traditional manipulator is based on serial design that has the advantages of large workspace and simple analysis, but suffers from many drawbacks such as low payload, low acceleration, low speed, high inertial, and accumulated error. The other kind of manipulators is based on parallel mechanisms which are expected to have those supplementary features: high payload, high acceleration, high speed, high rigidity (stiffness), low inertial, and non-accumulated error. These have potential to find important applications in areas like machining and automation in the semiconductor and electronic assembly industry. However, the analysis of the parallel manipulator is complicated because of the presence of constraints, and various singularities. As a result, so far, the available parallel platforms are not very popular and, indeed, have failed to outperform their serial counterparts impressively.; To solve the problems, in this research, we address several fundamental issues in the design and analysis of parallel manipulators. We use standard notions from differential geometry to study various spaces and the mapping involved. A general mathematical framework is established. We clarify concepts of singularities, study their effects to stiffness, dynamics and control, and propose methods to characterize and overcome them. We study and compare various dynamics formulations for parallel manipulators. Control algorithms for serial manipulators are successfully extended to parallel manipulators. For verification and comparisons, we apply our theory to a number of practical platforms in our lab for prototyping and experiments. In the setting of Riemannian Geometry, we propose the Generalized Newton's Law, which unifies different dynamics formulations for parallel manipulators. The control problem is naturally decomposed into the trajectory and constraint force control problems. Algorithms for both trajectory and constraint force control are devised. Finally we extend and apply the theory to the analysis of multifingered hand. |
