This paper focuses on the application of the intelligent control techniques (neural networks, fuzzy logic and genetic algorithms) anti their hybrid methods (neuro-fuzzy networks, neuro-genetic and fuzzy-genetic algorithms) in the area of humanoid robotic systems. This paper represents an attempt to give a report of the basic principles and concepts of intelligent control in humanoid robotics, with an outline of a number of recent algorithms used in advanced control of humanoid robots.

In this paper, a new method for selecting the appropriate compliance control parameters for robot machining tasks based on connectionist classification of unknown dynamic environments, is proposed. The method classifies the type of environment by using multilayer perceptron, and then, determines the control parameters for compliance control using the estimated characteristics. An important feature is that the process of pattern association can work in an on-line mode as a part of selected compliance control algorithm. Convergence process is improved by using evolutionary approach (genetic algorithms) in order to choose the optimal topology of the proposed multilayer perceptron. Compliant motion simulation experiments with robotic arm placed in contact with dynamic environment, described by the stiffness model and by the general impedance model, have been performed in order to verify the proposed approach.

In this paper, a new learning control algorithm based on neural network classification of unknown dynamic environment models and neural network learning of robot dynamic model is proposed. The method classifies characteristics of environments by using multilayer perceptrons, and then determines the control parameters for compliance control using the estimated characteristics. Simultaneously, using the second neural network the compensation of robot dynamic model uncertainties is accomplished. The classification capability of neural classifier is realized by efficient online training process. It is an important feature that the process of pattern classification can work in an online manner as a part of selected compliance control algorithm. Compliant motion simulation experiments have been performed in order to verify the proposed approach.

The paper addresses the problem of the practical stabilisation of manipulation robots which are in the contact with a dynamic environment. The model uncertainties represent the crucial problem in the control of robots within such tasks. Therefore, it is of practical interest to require a relaxed stability condition, the so-called practical stability of the system. A new method for analysis of a practical stability of robots is presented following the basic idea of a decomposition/aggregation method for the stability analysis of large-scale systems. The objective of the paper is to establish for the first time less conservative conditions for the practical stability of the robots around the programmed motion and interaction forces with the environment. The approach appears to be particularly suitable for the analysis of the so-called decentralised control laws. The elaborated procedure is demonstrated on an example.

The major concern of this paper is the application of connectionist architectures for fast online learning of robot dynamic uncertainties which are used at the executive hierarchical control level in the case of robot contact tasks. The connectionist structures are integrated in the nonlearning control laws for contact tasks which enable simultaneous stabilization and good tracking performance of position and force. It has been shown that the problem of tracking a specified reference trajectory and specified force profile with a preset quality of their transient response can be efficiently solved by means of application of the four-layer perceptron. The four-layer perceptron as part of hybrid learning control algorithms through the process of synchronous training use fast learning rules and available sensor informations in order to improve robotic performance progressively for minimal possible number of learning epochs. Some simulation results of deburring process with robot MANUTEC r3 are shown to verify effectiveness of the proposed control learning algorithms.

A major objective in this paper is the application of connectionist architectures for fast and robust online learning of dynamic relations used in robot control at the executive hierarchical level. The proposed connectionist robot controllers as new feature use decomposition of robot dynamics in the space of internal robot coordinates. In this way, this method enables the training of neural networks on the simpler input-output relations with significant reduction of learning time. The proposed controller structure comprises a form of intelligent feedforward control in the frame of decentralized control algorithm with feedback-error or driving torque error learning method. The another important features of these new algorithms are fast and robust convergence properties because the problem of adjusting the weights of internal hidden units is considered as a problem of estimating parameters by recursive least square method and Kalman filter approach. From simulation examples of robot trajectory tracking it is shown that when a sufficiently trained network is desired, the learning speed of the proposed algorithms is faster than that of the standard back propagation algorithm.>

A major objective in this paper is the application of new connectionist structures for fast and robust online learning of internal robot dynamic relations used as part of impedance control strategies in the case of robot contact tasks. Using proposed connectionist structures, stabilization of robot motion and interaction force with environment is achieved. The proposed neural network models with their special topology are integrated in position-based impedance control, force-based impedance control and stabilizing impedance control. In this way, efficient dynamic compensation and fast learning properties of the control algorithm for contact tasks are enabled. The effectiveness of the learning method is shown by simulation experiments of robot deburring process.>

In this paper the application of the damped least-squares method to the resolved-acceleration control is experimentally examined on a 2-DOF planar manipulator In order to decrease the position error introduced by the damping, only small singular values are damped. The symbolical expressions of the singular value decomposition of the Jacobian matrix were utilized in order to decrease the computational burden. Position error along the work-space boundary was only 15% greater than along the trajectories inside the reachable workspace.>

The application of connectionist architectures for fast and robust online learning of dynamic relations used in robot control at the executive hierarchical level is discussed. The proposed connectionist robot controllers use decomposition of robot dynamics. This method enables the training of neural networks on the simpler input/output relations with sigfnificant reduction of learning time. The other important features of these algorithms are fast and robust convergence properties because the problem of adjusting the weights of internal hidden units is considered as a problem of estimating parameters by the recursive least squares method and the extended Kalman filter approach. From simulation examples of robot trajectory tracking it is shows that when a sufficiently trained network is desired, the learning speed of the proposed algorithm is faster than that of the traditional backpropagation algorithms.>

The authors present two novel parallel algorithms for computing robot inverse dynamics and control laws starting from the customized symbolic robot models. The first algorithm resolves the scheduling problem for an array of pipelined processors. The second algorithm is devoted to parallel processors connected by a complete crossbar interconnection network. The main feature of the algorithms is that they take into account the communication delays between processors and minimize both the execution time and the communication cost. The algorithms were verified by experiments on an INMOS T800 transputer-based system. The experimental results showed that the most complicated dynamic control laws could be executed in a submillisecond range.>

A control strategy of manipulation robots in contact tasks such as robot cutting, surface grinding, polishing, and deburring is proposed. A digital force regulator of minimal variance has been applied in the case of an unknown, exterior system disturbance (such as workpiece surface roughness) during the robot cutting process. This regulator has been used to minimize an output signal variance and control signal oscillation. Stability conditions of the closed-loop control system are derived and discussed. A position/velocity controller is synthesized in Cartesian coordinate space to enable trajectory tracking during a desired technological operation. Based on simulation results some advantages of this control algorithm are outlined.>

The structure of SYM, a program package for manipulator modeling, control law synthesis, and simulation, is described. SYM's research and educational aims are emphasized. The control law synthesis in symbolic form and the system simulation are discussed in detail. Several examples of SYM outputs which depict the main steps in the manipulator control system creation process, namely, manipulator structure as a 3-D scheme, control law definition form, and system simulation results as 2-D plots are presented.>

An algorithm is presented for generating numerically efficient manipulator models in symbolic form, i.e., as computer programs written in a high-level language such as C or FORTRAN. The algorithm does its job in two main steps. First, a suboptimal symbolic model, which is the input for the second step, is generated. The first step has been explained in the authors' previous papers (1989). The main emphasis in this study is put on the second step: generating analytical expressions of output variables in the form of so-called structural matrices and their reconstruction into a program again.>

The authors address the problem of the optimal evaluation of robot inverse dynamics on array processors. The inverse dynamics models used are the symbolic customized models with near-minimum numerical complexity, which are computer-generated given the robot arm parameters. Such models represent the input for a proposed scheduling algorithm that distributes the computation of the model over several multipliers and adders. The scheduling algorithm is automatic and minimizes the number of microcycles. The algorithm was tested on several standard robots, and processor efficiency up to 84% was achieved. Experimental results on a 30 MFLOPS array processor showed that the inverse dynamics of a three-link PUMA-like robot requires 25.5 mu s. For a six-link robot, the computation of the inverse dynamics and the control law takes about 100 mu s on this processor. Thus, the control of high-speed robots could be achieved by attaching low-cost array processors (10 MFLOPS) to the control processor of the robot controller.>

An algorithm for constructing dynamic models of single-arm robots is presented in this paper. Motion equations of robots in analytical form are derived applying a fully automated procedure. It is shown that the solution of direct and/or inverse problem based on the analytical model requires considerably fewer floating- -point multiplications/additions that is the case with previously developed numerical methods. The developed method is therefore very suitable for real-time application of the dynamic models of robots. The developed program package is illustrated using the example of Stanford manipulator.

The solution to the inverse manipulation problem for redundant manipulators has mostly been considered from a geometric-kinematic standpoint. A procedure for the inverse problem solution using the dynamic model of the manipulator and its actuators is developed. Nominal trajectories in the space of joint coordinates are generated so as to be optimal with respect to total energy consumption of the actuators (hydraulic or electric). The treatment of constraints on joint coordinates and rates is also involved in the procedure. The algorithm is illustrated by two industrial robots.

Based on a comparison of the computational complexity of different spatial mechanisms dynamics formulations it is shown that the most appropriate one is the Newton-Euler formulation using recurrence relations for velocities, accelerations, and generalized forces. This is based on numerical efficiency and intermediate results. A general algorithm which solves both the direct and inverse problem of dynamics for an open-chained spatial mechanism of an arbitrary mechanical configuration is developed and realized. It is pointed out that for control algorithms which assume knowledge of system dynamics in real time, it is necessary to compute the inertial matrix and the term taking into account the rest of the dynamical effects separately. The "accelerated" computational algorithm for real-time implementation with this property has been developed and realized. Up to now reported real-time computational schemes do not have this feature directly implementable. Depending on the control law, manipulator configuration, type of functional tasks, and given ranges of operational speed it can appear that it is sufficient to compute only the dominant dynamical influences and not the complete dynamics. The criterion for the optimal choice of the level of the approximation of the dynamical model is formulated based on nominal regimes for specific functional tasks. For this purpose the algorithm for generation of the mathematical model of variable complexity is developed and realized. The procedure is implemented on two typical manipulator mechanical configurations of six degrees of freedom (d.o.f.) and a comparison of exact and various approximate models is performed.

A procedure for the interactive computer-aided control synthesis of manipulators is presented. The procedure is based on an approach to control synthesis in which the dynamics of the system are taken into account, but the resulting control law is simple and applicable. All phases of the procedure are briefly described. At each step of the control synthesis, interaction with the user is emphasized underlining the ability of the procedure to include user experience. The algorithm is applied to a six degree of freedom industrial manipulator.

This paper describes, in a polemical way, the possible ways for synthesis of artificial gait, taking into account its application in the rehabilitation of severely disabled paraplegics. The essential relation between the hierarchical control concept of artificial legged locomotion systems and human beings has been treated. The method of prescribed synergy has been described in detail. This method, in the author's opinion, renders wide possibilities in solving delicate problems in the synthesis and control of artificial anthropomorphic gait. The possibility of applying the gait-logic description in the synthesis of finite automata and the possibility of driving system hybridization in the rehabilitation of persons with muscular insufficiency have also been considered.