An Improved Hestenes SVD Method and Its Parallel Computing and Application in Parallel Robot

Singular value decomposition (SVD) of matrix is an important and familiar problem in maths science and engineering. Among many SVD methods, Hestenes method is widely used as it is suiting for parallel processing in particular. An improved Hestenes SVD method is proposed in this paper, which notably reduces the sweep numbers and orthogonalization numbers during the matrix singular value decomposition. It also facilitates and quickens the process of computing (generalized) inverse matrix. In addition, two kinds of parallel algorithms are studied for the improved Hestenes SVD method based on row and column division respectively, and then their performance and efficiency are analyzed. Influence coefficient plays an important role in the analysis of parallel robot's kinematics and dynamics, and the second-order influence coefficient matrix shouldn't be ignored especially in the condition of high speed. Aiming at the characteristics of parallel robot's increasing computing requirement, the experiments about the improved Hestenes SVD method and its parallel algorithm are done by computing the first-order and second-order influence coefficient matrix of 6-DOF parallel robot. Experiment results show that the proposed method can improve computing efficiency greatly, and be beneficial to parallel robot's kinematics, dynamics performance analysis and real time control based on lots of influence coefficient matrix computing. The proposed method also suits for many other engineering fields with similar matrix processing.