Neural Network Based Tuning Algorithm for MPID Control

Abstract

One of the biggest problems for space manipulators are to cope with flexibility. If manipulator links undergo deflection during the course of operation, it may prove difficult to reach a desired position or to avoid obstacles. Furthermore, once the manipulator has reached a set point, the residual vibration may degrade positioning accuracy and may cause a delay in task execution. At the same time the flexible manipulators has the advantages of high payload to weight ratio, which make them superior in the space exploration and orbital operation. The high payload to weight ration is not the only merits of using flexible manipulators in space application. Lower power consumption, smaller actuators and speedy operation make the flexible manipulators the optimum choice for space manipulators. Since Cannon et al. (Cannon & Schmitz, 1984) started initial experiments using the linear quadratic approach methods to control flexible link manipulators, much researches on the usage of flexible manipulator had been developed. Using the approach of enhancement the measurements of the vibration variables was studied by (Ge et al., 1999; Sun et al., 2005) while Etxebarria et al. (Etxebarria et al., 2005) gives attention to the algorithms used in controlling the flexible manipulators. The enhancement of the traditional PD controller by adding a vibration control term is one of the most effective methods for the flexible manipulators. Lee et al. proposed PDS (proportional-derivative strain) control for vibration suppression of multi-flexible-link manipulators and analysed the Liapunov stability of the PDS control (Lee et al., 1988). Maruyama et al. (Maruyama et al., 2006) developed a golf robot whose swing simulates humanmotion. They presentedmodel accounting for golf club flexibility with all parameters identified in experiments and generated and implemented trajectories for different criterion such as minimizing total consumed work, minimizing summation of the squared derivative of active torque andmaximizing impact speed. Matsuno and Hayashi applied the PDS control to a cooperative task of two one-link flexible arms (Matsuno & Hayashi, 2000). They aimed to accomplish the desired grasping force for a common rigid object and the vibration absorption of the flexible arms. A neural network is a data modelling tool that is able to capture and represent complex input/output relationships. The motivation for the development of neural network technology stemmed from the desire to develop an artificial system that could perform “intelligent tasks similar to those performed by the human brain. Neural Networks resemble the human brain in the following two ways: 8