Acceleration Of A Large-scale Interior-point Based Inverse Kinematics And Its Applications

Hyper-redundant manipulators has ideal fault-tolerance and maximal tip load for outer-space exploration.Due to computational complexity in the inverse kinematics,IK solver for such manipulator is considered the most tough problem in this area,and the further application is also hindered.The IK problem is be approached using a fast interior point method by transforming the IK to a quadratic constrained quadratic programming problem.The advantages of fast convergence from interior point can be well utilized in the area of interests.Exploitation on sparsity of the problem is also made in accelerating an interior point method by viewing the problem from the perspective of scientific computing.The linear solver is based on Cholesky decomposition and is accelerated on a CPU-GPU system.The profiling results indicates a significant performance boost especially for large scale problems.Next simulations are conducted on a 2 dimensional parallel manipula?tor,a hyper-redundant variable geometry trussarm and a binary manipulator.To visualize the robotic arm,a simulation platform using 3Ds Max was built,which consists of three modules for algorithm,UI and visual-ization respectively.For a continuous VGT,three groups of simulation under both normal and fault condition were conducted.Results indicated a fast solving capability in our algorithm.Trajectory planning under fault condition is also tested.The computation cost is small and fault tolerance is also well utilized.Further,a trajectory simulation is conducted as in each step the initial position is selected as the results obtained from the previous one.Minimized changes in the lengths of actuators are observed.Finally,the algorithm for a binary manipulator was implemented,a multi-objective optimization approach was adopted to transform the binary constraints to continuous objective function.Desired results are obtained which also offers great insights for future studies.