This paper presents a computational scheme on a pipelined vector processor for robot inverse dynamics. Vectorization of the inverse dynamics computation is achieved while the dynamics formulation still remains within the general linear recursive framework. Parametric studies are conducted on a commercially available pipelined vector processor. The 1.82 speedup is achieved through vectorization for the inverse dynamics computation of a general six-degrees-of-freedom robot manipulator.>

A computational scheme for a pipelined vector processor for robot inverse dynamics is presented. Vectorization of the inverse dynamics computation is achieved while the dynamics formulation still remains within the general linear recursive framework, which is based on Newton-Euler equations and originally designed for scalar processing. Parametric studies were conducted on a commercially available pipelined vector processor. It is shown that the factor 1.82 of the speedup is achieved through vectorization for the inverse dynamics computation of a general six-degree-of-freedom robot manipulator.>

It is shown that the principle of superposition can be used to analyze the spatial motion of a bevel-geared three-roll wrist with nonzero oblique angle. The total motion of the wrist is decomposed into three independent component motions, which are easier to analyze. A superposition tabulation is then constructed to derive the relations among the three wrist-joint variables and three coaxial actuation variables. The approach is straightforward and can be adapted for analyzing the motion of other geared wrists that are variations of the wrist considered.>

Robot manipulators have a variety of wrist arrangements. In commonly found wrist arrangements, the wrist axes are either cointersecting or nearly cointersecting, but rotations at some wrist joints must be restricted to avoid mechanical interference. Closed-form algebraic solutions for many simple robots with such wrists have been developed over the years. This paper deals specifically with robots with continuous 3-roll wrists. In a continuous 3- roll wrist, all wrist joints have unrestricted spin freedom without any mechanical inference. Although several manipulators now exist with such wrists, hardly any literature exists which deals with them. We present explicit closed-form solutions for robots with two types of continuous 3-roll wrists. The zero reference position method has been used to derive these solutions. The results presented in this paper should be useful in developing computer control for these robots with continuous 3-roll wrists.

The zero reference position method for analyzing robot motions is based upon an arbitrary choice of the zero reference position, the principle of displacement similarity, and the hand-to-the base sequence of changing the joint variables in a serial manipulator. Its applications in performing recursive computations, formulating explicit governing equations, and obtaining closed form solutions for serial manipulators are presented in this paper. It is shown to be an efficient method for updating manipulator positions. A Bendix type of industrial robot is used for illustrating the closed form solution procedure.

A simplified description of robotic manipulator is in terms of its zero reference position. It requires the specification of the joint axes directions and the coordinates of points locating the joint axes in the base coordinate system. Manipulator dynamics is developed in an extended zero reference position description. The recursive Newton-Euler formulations for the problems of inverse and direct dynamics are presented.

Numerical solution of the inverse kinematics of robots using modified Newton-Raphson (MNR) and modified predictor-corrector (MPC) algorithms is discussed. These modified algorithms are highly reliable and stable. Both algorithms always find a solution if a physically realizable robot configuration exists. They are also capable of approaching singular configurations of the robot as E = C εm, where E is the error between the theoretical and numerically computed singular joint values,\epsivis the convergence criteria, and C and m are appropriate constants. It is also shown that the modified predictor-corector (MPC) algorithm is 5-15 times faster than the modified Newton-Raphson (MNR) algorithm.