Task-optimized cable-actuated planar parallel manipulator architecture and its concurrent implementation

Abstract

This work presents an initial framework for an efficient new technique for obtaining task-optimized parallel manipulators with the aid of parallel computing through OpenMP directives. A cable-driven parallel manipulator is an architecture whose actuated limbs are cables. All of the cables must remain in constant positive tension to constrain the motion of the moving end-effector. A Differential Evolution algorithm is applied in order to optimize the topology and actuator specifications of a cable-driven parallel manipulator. The algorithm's intrinsic parallelism is exploited using OpenMP directives to evaluate the manipulator's associated reachable and wrench workspaces. The results show that this algorithm is effective at obtaining a task-optimized architecture for the cable-driven parallel manipulator. Parallel implementation is shown to improve the algorithm's performance with a speedup of 7.4 times using ten cores on the Atlantic Computational Excellence Network (ACEnet) Fundy compute resource which utilizes Parallel Sun x4600 and x2200 AMD Opteron (dual-core) clusters.