This study is an effort to give a practical solution in the problem of optimizing the structure of the Hierarchical Mixture of Experts model, which is a natural extension of the Associative Gaussian Mixture of Experts...
详细信息
ISBN:
(纸本)0780382781
This study is an effort to give a practical solution in the problem of optimizing the structure of the Hierarchical Mixture of Experts model, which is a natural extension of the Associative Gaussian Mixture of Experts system. We present two novel methods for optimizing such structures using Genetic Algorithms. Special concern is taken for reducing the computational time so as to efficiently allow the structure to "grow" while it evolves with the Genetic Algorithm. The main contribution of the paper lies on the efficient, topologically oriented, representations of such architectures so as to be optimized through involving genetic algorithms.
We describe the latest additions to SOSTOOLS, a freely available MATLAB toolbox for formulating and solving sum of squares programs. Among the many improvements, there are native polynomial objects, structure-exploiti...
详细信息
ISBN:
(纸本)0780383354
We describe the latest additions to SOSTOOLS, a freely available MATLAB toolbox for formulating and solving sum of squares programs. Among the many improvements, there are native polynomial objects, structure-exploiting techniques for sparse and structured polynomials, new customized functions, and support for alternative SDP solvers. We sketch some of the theory behind the new improvements, and illustrate the new commands using control-oriented examples.
Multi parameteric quadratic programming gives a full offline solution to a time varying quadratic programming (QP) problem arising during constrained predictive control. However, coding and implementation of this solu...
详细信息
Multi parameteric quadratic programming gives a full offline solution to a time varying quadratic programming (QP) problem arising during constrained predictive control. However, coding and implementation of this solution may be more burdensome than simply solving the original QP. This paper proposes an algorithm, which by accepting a small amount of suboptimality in the predicted control trajectories, achieves a large decrease in both the online computation and data storage requirements
The possibility of operating in remote environments by means of teleoperated systems has always been considered of relevant interest in robotics. For this reason, in this paper, the relationship between a slave robot ...
详细信息
ISBN:
(纸本)0780384636
The possibility of operating in remote environments by means of teleoperated systems has always been considered of relevant interest in robotics. For this reason, in this paper, the relationship between a slave robot and the uncertain remote environment is proposed as the impedance to generate the virtual force to feed back to the operator. For the control of a teleoperated mobile robot equipped with camera, the teleoperated mobile robot take pictures of remote environment and sends the visual information back to the operator over the Internet. Because of the limitation of communication bandwidth and narrow view-angles of camera, it is not possible to watch the environment clearly, especially shadow and curved areas. To overcome this problem, the virtual force is generated according to both the distance between the obstacle and robot and the approaching velocity of the obstacle. This virtual force is transferred back to the master over the Internet and the master, which can generate force, enables a human operator to estimate the position of obstacle in the remote environment. By holding this master, in spite of limited visual information, the operator can feel the spatial sense against the remote environment. This force reflection improves the performance of a teleoperated mobile robot significantly.
The developments and control applications of SOSTOOLS, a free third-party MATLAB toolbox for formulating and solving sum of squares programs was discussed. The technique was based on the sum of squares decomposition f...
详细信息
The developments and control applications of SOSTOOLS, a free third-party MATLAB toolbox for formulating and solving sum of squares programs was discussed. The technique was based on the sum of squares decomposition for multivariate polynomials which can be effectively computed using semidefinite programming. The nonlinear stability analysis, parametric robustness analysis, safety verification, and nonlinear controller synthesis was also considered in the control applications. A sparsity structure was used while considering infinitely constrained linear matrix inequalities (LMI).
Much of the work on predictive based multi-rate control has been based on the GPC algorithm (Clarke et al, 1987). However academic practitioners in single rate predictive control tend to favour approaches with better ...
详细信息
Some kinds of speed control loops can be easily designed based on simple mechanical dynamics by using reasonable assumptions in AC motor controlsystems. However, these simple approaches give undesired performances un...
详细信息
Some kinds of speed control loops can be easily designed based on simple mechanical dynamics by using reasonable assumptions in AC motor controlsystems. However, these simple approaches give undesired performances under mismatches of model parameters, external load conditions and unknown dynamics. Moreover, in real systems as well as during analysis, speed estimation error is inevitable and should be considered from the viewpoint of the stability of the whole feedback controlsystem. To cope with the robust characteristics under inherent uncertainties, a nonlinear uncertainty observer using the radial basis function networks (RBFNs) is proposed. A control law for stabilising the system and adaptive laws for updating both weights in the RBFNs and a bounding constant are established, so that the whole closed-loop system is stable in the sense of Lyapunov. Additionally, stability proof of the whole controlsystem including the speed estimation error is presented. The proposed approach is applied to a three-level inverter-fed induction motor direct torque control (DTC) system, and computer simulations as well as experimental results are presented to show the validity and effectiveness of the proposed system.
This paper investigates the utilisation of neural networks (NNs) and parametric linear approaches for modelling a twin rotor multi-input multi-output system (TRMS) in hovering position. A multi-layer perceptron (MLP) ...
详细信息
This paper investigates the utilisation of neural networks (NNs) and parametric linear approaches for modelling a twin rotor multi-input multi-output system (TRMS) in hovering position. A multi-layer perceptron (MLP) neuro-model is designed to characterise the TRMS. A parametric model of the system is then developed with the conventional recursive least square (RLS) technique. A comparative assessment of the two model types, in characterising the system, is carried out in the time and frequency domains. Experimental results demonstrate the superiority of the NN approach over the conventional linear modelling approach. The developed neuro-modelling approach will be used for control design and development in future work.
For a hysteresis structure with energy dissipation devices, the force-displacement relation is nonlinear such that it is very difficult to evaluate the actual damping and stiffness coefficients, even if the force-disp...
详细信息
This paper investigates the utilisation of back propagation neural networks (NNs) for modelling flexible beam structures in fixed-free mode; a simple representation of an aircraft wing or robot arm. A comparative perf...
详细信息
This paper investigates the utilisation of back propagation neural networks (NNs) for modelling flexible beam structures in fixed-free mode; a simple representation of an aircraft wing or robot arm. A comparative performance of the NN model and conventional recursive least square scheme, in characterising the system is carried out in the time and frequency domains. Simulated results demonstrate that using NN approach the system is modelled well than with the conventional linear modelling approach. The developed neuro-modelling approach would further be utilized in the design and implementation of suitable controllers, for vibration suppression in such system.
暂无评论