This paper is concerned with the design of linearset -- point controllers for a specific nonlinear system: a motor driven pendulum. It is shown that when input constraints and delays are present, the synthesis procedure is non trivial. Even if the design is based on mature theoretical basis, extensive nonlinear simulations are the only practical way to ensure that the linear design offers acceptable tracking performance in a large operation range while respecting the requirements of input constraints and robustness.

Flexible and dexterous manipulation of modern industrial applications requires the use of robots with extra (more than six) axes of motion (degrees of freedom). Thus the kinematic and control considerations of redundant robots are currently of increasing interest to both theorists and practitioners. For this type of robots, a closed-form solution of the inverse kinematic problem is not possible, and the classical approach is to use the pseudoinverse of the Jacobian matrix together with an extra criterion function.This paper presents an iterative technique for velocity control, which is suitable for redundant robots with a maximum of 11 d.o.f. This technique, which splits the Jacobian inversion problem into two subproblems (wrist position and wrist orientation), provides an exact solution for the arm tip position and an approximate solution for the wrist orientation. The convergence of the procedure is examined and a simulated example is included.

A recent discrete-time robotic model is employed and linearized to develop three different LQ-type robot controllers. These controllers are: optimal time-varying linear-quadratic controller (OTVLQ), OTVLQ controller with steady-state error elimination, and one-step ahead optimal LQ controller. These controllers require moderate computational effort and are offered for microprocessor-based implementation. Experimental results on a KUKA IR 160/15 robotic model verified the success and efficiency of the controllers.

Flexibility and versatility are two basic requirements for industrial robots used in manufacturing areas. In the present paper, a kinematic transformation for redundant robots between Cartesian and joint coordinates is achieved in the trajectory planner. This problem is treated by taking into account both the dynamics of the system expressed by its Lagrangian and the kinematic constraints of the robot. Given the joint trajectories as above, a control law for joint trajectory tracking is generated here using the Model Based Predictive Control (MBPC) approach. The MBPC strategy is based on an explicit robot model to predict the process output over a long-range time period. The postulated control law is constructed such that to fit a desired command acceleration for the servo motors of the robotic system. The actual control law is calculated by minimizing a suitable objective function. A numerical example concerning a robot with one degree of redundancy illustrates the method and shows the excellent performance of the proposed MBPC scheme. The drifting problem that emerges in the resolved motion control of redundant robots is successfully faced.

Part I of this paper presents a study of an autonomous trajectory generation system (ATGS) which is useful for industrial robots carrying out mechanical operations. It is a very simple job with such an ATGS to generate planar curved trajectories f(x, y) = 0, an attractive factor in such applications. Two alternative variations of the ATGS are presented which possess particular characteristics. Part II deals with the trajectory control problem and uses the ATGS as a first stage of a model reference adaptive control loop of a robot. The theoretical results are illustrated by simulation examples worked out using the software packages developed for this purpose.

A method for controlling a 6-degrees-of-freedom robotic manipulator based on a decomposition of its full dynamical model in two 3-degrees-of-freedom submodels (one for the lower part and one for the upper part) is presented. The lower-part (arm) submodel takes into account the effect of the upper part (wrist) and of the robot task requirements, in the form of an external force/torque pair expressed in tool coordinates. The influence of the robot task is similarly included in the wrist dynamic model. The control algorithm is of the model reference adatpvie (MRAC) type which is actually a nonlinear proportional plus integral (PI) algorithm. Based on previous results a simplified MRAC controller is derived which needs less computational effort for its tuning. As a simulator of the 6-degrees-of-freedom manipulator, the exact dynamic Newton-Euler model is used. The paper includes a number of computational experimental results which show the effectiveness of the method.

The complete dynamic model of a 6-degrees-of-freedom (DOF) robotic manipulator is very complicated, having a large number of nonlinear terms, and so it is very difficult to be used for actual on-line control. The purpose of this paper is to provide a method for decomposing the dynamical model of such a 6-IDOF manipulator in two 3-DOF submodels (one for the arm and one for the wrist) which can be successfully used both for simulation and control purposes. The arm submodel takes into account the effect of the wrist and of the manipulator task requirements, in the form of an external force/ torque pair expressed in tool coordinates. The influence of the robot task is similarly included in the wrist dynamic model. Then the paper shows how the model reference adaptive control (MRAC) technique can be used on this two-part robot model. As a simulator of the 6-DOF robot, the exact dynamic model, based on the Newton Euler method, is employed. Computational experimental results are included which show the effectiveness of the approach.