الفهرس | Only 14 pages are availabe for public view |
Abstract In this thesis, three different planar parallel manipulators (3-RRR, 3RPR, 3- PRR) kinematics were studied. First, the direct kinematics was solved for the three manipulators under study. Secondly, two models of artificial neural networks with multiple neural networks were proposed, namely, multilayer perceptrons (MLP) and selforganizing maps (SOM) to determine the unique solution, the actual physical solution, in among all possible solutions result from the kinematics model. MLP has better performance with acceptable prediction errors with less training time required compared to SOM. Thirdly, two different approaches were proposed for trajectory tracking control of a planar parallel manipulator with singularity avoidance. The first approach depends on allowing the platform orientation to change by constant steps under the control scheme. The second approach depends on allowing the platform orientation to change by a certain angle calculated by minimizing the actuating forces. The proposed approaches can be easily extended to any class of parallel manipulators either planar or spatial. |