摘要:

This paper describes the generation of closed chain (CC) dynamics for robotic manipulators based on models of the open chain (OC) kinetics and environmental dynamics. The motivation for this exercise is to control the manipulator motion as well as the interaction forces at the end-effector. A fixed gain controller (the Trajectory-Following Proportional/Integral/Derivative—or TFPID— controller), is synthesized from the CC dynamics, and this performs adequately. However, the prime objective is to describe an alternative route via the adaptive Minimal Control Synthesis (MCS) algorithm (Stoten 1990, Stoten and Hodgson 1991). This control strategy requires no complicated OC, CC or environmental models in its formulation. It is shown in implementation studies that MCS produces excellent force and motion trajectory tracking, which is far superior to that resulting from the TFPID controller.

ISSN:0020-7179    

DOI:10.1080/00207179208934325

出版商:Taylor & Francis Group    

年代:1992

数据来源: Taylor

摘要:

In this paper, we consider the deadbeat control for both the output and input of step-like references and single-input single-output processes. The controllers are selected from among two-degree-of-freedom robust-tracking controllers, and we derive the parametrization of all input/output deadbeat controllers. This parametrization is given by solving two polynomial Diophantine equations. We also derive the parametrization of all references for which input/output deadbeat response is maintained. In addition, we investigate the conditions for the processes under which the output deadbeat always implies the input deadbeat.

ISSN:0020-7179    

DOI:10.1080/00207179208934326

出版商:Taylor & Francis Group    

年代:1992

数据来源: Taylor

摘要:

This paper presents a comparative study of the performance of a multi-rate output controller (MROC) with a linear quadratic gaussian (LQG) controller in the presence of noise disturbances and anti-aliasing filters. The basis of comparison is to apply an LQG law with a one-step-ahead prediction-type Kalman filter (hereafter called LQG law I) and an LQG law with a current estimation-type Kalman filter (hereafter called LQG law II) to a linear time-invariant continuous-time plant model with a white gaussian process and measurement noise and compute a linear quadratic performance index for the discretized plant. Equivalent noise matrices for using the MROC law are derived and the same quadratic performance index computed. In order to have a fair comparison, the cut-off frequency of the anti-aliasing filter used to remove high frequency noise components prior to sampling is also kept the same when applying both laws. Application of both laws in typical situations shows that the performance of the MROC law is worse than either that of LQG law I or LQG law II.

ISSN:0020-7179    

DOI:10.1080/00207179208934327

出版商:Taylor & Francis Group    

年代:1992

数据来源: Taylor

摘要:

The problem of making the Kalman filter robust is considered in the paper. Proceeding from the equivalence between the Kalman filter and the least squares regression problem, a statistical approach namedM-estimation is suggested to resolve the regression problem robustly. Since the derived robustM-filters do not have an attractive recursive form, the possibility is proposed of designing real-time estimators based on the general formulation of the robust stochastic approximation algorithm and step-by-step optimization with respect to the weighting matrix combined with suitable approximations. Results of simulation demonstrating the robustness of the proposed estimators are also included.

ISSN:0020-7179    

DOI:10.1080/00207179208934328

出版商:Taylor & Francis Group    

年代:1992

数据来源: Taylor

摘要:

The not previously considered problem of decoupling time varying multivariable systems is investigated and two dual-purpose algorithms, which decouple and regulate the system, are derived. One algorithm also gives pole assignment and the other minimum variance control. The algorithms are shown to stabilize and decouple the closed loop system.

ISSN:0020-7179    

DOI:10.1080/00207179208934329

出版商:Taylor & Francis Group    

年代:1992

数据来源: Taylor

摘要:

A new set of stability tests for differential non-unit memory linear repetitive processes is developed using plant step response data. These tests then lead, at no extra computational cost, to unique information concerning two crucial aspects of the behaviour of these processes. The benefits of this information in the formulation and solution of practically feasible control policies are clearly shown by the initial results on this general area reported here. Finally, some open research problems are noted.

ISSN:0020-7179    

DOI:10.1080/00207179208934330

出版商:Taylor & Francis Group    

年代:1992

数据来源: Taylor

摘要:

Simplex type dynamic output feedback variable structure control for stabilizing multivariable linear time-invariant plants is analysed. Two types of controllers are considered; a compensator type and an observer type. It is shown that while the compensator-type controller stabilizes the plant only if the plant is minimum-phase, the observer-type controller does not face the same restriction. The design of low-order (less than the order of the plant) observer-type controllers is also considered.

ISSN:0020-7179    

DOI:10.1080/00207179208934331

出版商:Taylor & Francis Group    

年代:1992

数据来源: Taylor

摘要:

A robotic force controller is usually designed in terms of an exact system model, such as the modified computed torque controller developed by McClamroch and Wang. If the mathematical model of the robot system is exact, the modified computed torque can simultaneously control the robot motion and the contact force in an accurate way. However, there may exist uncertainties in the model, such as the flexibility of joints and links, the joint friction, and an inexact surface model. It will be shown that the modified computed torque controller may result in an unstable closed-loop system for a system with uncertainties. This difficulty can be overcome by introducing a robust controller. Two robust force controllers are proposed. One is based on the non-singular model system so that the force is indirectly controlled. The other is based on the transformed system model so that both the robot motion and the force can be simultaneously controlled. The controllers are robust in that they are insensitive to variations in the plant parameters and to external disturbances of contact force.

ISSN:0020-7179    

DOI:10.1080/00207179208934332

出版商:Taylor & Francis Group    

年代:1992

数据来源: Taylor

摘要:

The ability of a neural network to represent an input-output mapping is usually only measured in terms of the data fit according to some error criteria. This ‘black box’ approach provides little understanding of the network representation or how it should be structured. This paper investigates the topological structure of multilayer feedforward neural networks (MFNN) and explores the relationship between the numbers of neurons in the hidden layers and finite dimensional topological spaces. It is shown that a class of three layer (two hidden layer) neural networks is equivalent to a canonical form approximation of nonlinearity. This theoretical framework leads to insights about the architecture of multilayer feedforward neural networks, confirms the common belief that three layer (two hidden layer) feedforward networks are sufficient for general application and yields an approach for determining the appropriate numbers of neurons in each hidden layer.

ISSN:0020-7179    

DOI:10.1080/00207179208934333

出版商:Taylor & Francis Group    

年代:1992

数据来源: Taylor

摘要:

We consider worst-case l1identification of causal linear shift-invariant systems from time series. Many results are given on general aspects of identification algorithm performance, existence of optimal algorithms, robust convergence, and input (experiment) design. The identification methodology studied here is compatible with the modelling requirements of modern robust control design.

ISSN:0020-7179    

DOI:10.1080/00207179208934334

出版商:Taylor & Francis Group    

年代:1992

数据来源: Taylor