No abstract

a b s t r a c tSingularities and uncertainties in arm configurations are the main problems in kinematics robot control resulting from applying robot model, a solution based on using Artificial Neural Network (ANN) is proposed here. The main idea of this approach is the use of an ANN to learn the robot system characteristics rather than having to specify an explicit robot system model.Despite the fact that this is very difficult in practice, training data were recorded experimentally from sensors fixed on each joint for a six Degrees of Freedom (DOF) industrial robot. The network was designed to have one hidden layer, where the input were the Cartesian positions along the X, Y and Z coordinates, the orientation according to the RPY representation and the linear velocity of the end-effector while the output were the angular position and velocities for each joint, In a free-of-obstacles workspace, off-line smooth geometric paths in the joint space of the manipulator are obtained.The resulting network was tested for a new set of data that has never been introduced to the network before these data were recorded in the singular configurations, in order to show the generality and efficiency of the proposed approach, and then testing results were verified experimentally.

This paper is devoted to the development and implementation of neural network technology to solve the inverse kinematics problems for serial robot manipulators, given the desired Cartesian path of the end effector of the manipulator in a free-of-obstacles workspace. Offline smooth geometric paths in the joint space of the manipulator are obtained. The proposed technique does not require any prior knowledge of the kinematics model of the system being controlled; the main idea of this approach is the use of an artificial neural network to learn the robot system characteristics rather than having to specify an explicit robot system model. Since one of the most important problems in using artificial neural networks is the choice of the appropriate network configuration, two different configurations were compared; they were trained to learn the desired set of joint angles positions from a given set of end effector positions. The generality and efficiency of the proposed algorithm are demonstrated through simulations of a general six-degrees-of-freedom serial robot manipulator.

This paper discusses the use of artificial neural networks (ANNs) as a method of trajectory tracking control for a robotic system. Using an ANN does not require any prior knowledge of the kinematics model of the system being controlled; the basic idea of this concept is the use of the ANN to learn the characteristics of the robot system rather than to specify an explicit robot system model. In this approach, disadvantages of some schemes such as the fuzzy learning control, for example, have been elevated. Off-line training was performed for a geometric trajectory that is free of obstacles. Studying the kinematics Jacobian of serial manipulators by using ANNs has two problems: one of these is the selection of the appropriate configuration of the network and the other is the generation of a suitable training dataset. In this approach, although this is very difficult in practice, training data were recorded experimentally from sensors fixed on each joint to overcome the effect of kinematics uncertainties present in the real world such as ill-defined linkage parameters, links flexibility, and backlashes in the gear train. Then two network configurations were compared to find the best configuration to be used. Finally, the simulation results were verified experimentally using a general six-degree-of-freedom (DOF) serial robot manipulator.

An adaptive learning algorithm using an artificial neural network (ANN) has been proposed to predict the passive joint position of under-actuated robot manipulator. In this approach, a specific ANN model has been designed and trained to learn a desired set of joint angular positions for the passive joint from a given set of input torque and angular position for the active joint over a certain period of time. Trying to overcome the disadvantages of many used techniques in the literature, the ANNs have a significant advantage of being a model-free method. The learning algorithm can directly determine the position of its passive joint, and can, therefore, completely eliminate the need for any system modelling. Even though it is very difficult in practice, data used in this study were recorded experimentally from sensors fixed on robot's joints to overcome the effect of kinematics uncertainties present in the real world such as ill-defined linkage parameters and backlashes in gear trains. An ANN was trained using the experimentally obtained data and then used to predict the path of the passive joint that is positioned by the dynamic coupling of the active joint. The generality and efficiency of the proposed algorithm are demonstrated through simulations of an under-actuated robot manipulator; finally, the obtained results were successfully verified experimentally.