In this paper, a new method for the computation of the optimal step in gradient algorithms for a class of linear quadratic optimal control problems is presented. This method improves the convergence of the gradient al...
详细信息
In this paper, a new method for the computation of the optimal step in gradient algorithms for a class of linear quadratic optimal control problems is presented. This method improves the convergence of the gradient algorithms and is optimal, i.e., it out performs any other step searching scheme in the above class of problems.
In this paper the Supervisory control Problem (SCP) for discrete event systems (DES) is generalized to a class of probabilistic discrete event systems (PDES). Necessary and sufficient conditions for the existence of a...
详细信息
In this paper the Supervisory control Problem (SCP) for discrete event systems (DES) is generalized to a class of probabilistic discrete event systems (PDES). Necessary and sufficient conditions for the existence of a solution to the probabilistic SCP for a class of nonterminating PDES are developed. Methods of representing probabilistic supervisors for PDES are described and the computation of supervisors is briefly discussed. Finally, we discuss how the results may be extended to terminating probabilistic languages.< >
作者：
P.C. BreedveldElectrical Engineering Department
Control Systems and Computer Engineering Group Control Laboratory and Mechatronics Research Centre Twente University of Twente Enschede Netherlands
The author identifies incentives for the use of computer-aids in modelling and design of physical systems and discusses possible solutions.
The author identifies incentives for the use of computer-aids in modelling and design of physical systems and discusses possible solutions.< >
This paper derives the tight upper bound on the number of stationary points that an nth order Hopfield network can have, and provides a characterization of synaptic sign patterns that allow a network to achieve this c...
详细信息
This paper derives the tight upper bound on the number of stationary points that an nth order Hopfield network can have, and provides a characterization of synaptic sign patterns that allow a network to achieve this capacity. We show that n/sup C//sub [/n/2] is the tight upper bound on the number of stationary points of an n-order Hopfield network. A simple graph theoretic characterization of the maximum capacity networks is given.< >
This paper is concerned with the design of a discrete time Model Reference Adaptive control System (WACS) for motion control under unknown nonlinear friction. Static and Coulomb friction forces are represented by a de...
详细信息
This paper is concerned with the design of a discrete time Model Reference Adaptive control System (WACS) for motion control under unknown nonlinear friction. Static and Coulomb friction forces are represented by a deadzone element for the input, which is active at low velocities. and disturbance. which may take different values depending on the motion direction. respectively. Then, the WAes for the nonlinear friction is a cascade combination of a nonlinear block which contains the deadzone and the disturbance term and a linear dynamic block. Experimental results from a rotary servo system demonstrate the effectiveness of the proposed WAeS
作者：
Daniel E. RiveraSaurabh BhatnagarDepartment of Chemical
Bio and Materials Engineering and Control Systems Engineering Laboratory Computer-Integrated Manufacturing Systems Research Center Arizona State University Tempe AZ USA
A novel technique for identifying reduced-order models in the closed-loop is presented. The method arrives at a process model and its corresponding compensator in an iterative fashion by introducing a series of step c...
详细信息
A novel technique for identifying reduced-order models in the closed-loop is presented. The method arrives at a process model and its corresponding compensator in an iterative fashion by introducing a series of step chan at the manipulated variable. The bias introduced into the identification data set by the closed-loop system, coupled with a control-relevant prefilter, yields a model whose corresponding control system improves its performance at every step. The method is appealing to chemical engineering practitioners because it combines the tasks of system identification with controller commissioning to produce a simple-to-use yet reliable autotuning procedure.
The problem of improved performance adaptive control (IPAC) of a class of linear and nonlinear systems is considered. A method for its solution is presented, the main feature of which lies in augmenting the "stan...
详细信息
The problem of improved performance adaptive control (IPAC) of a class of linear and nonlinear systems is considered. A method for its solution is presented, the main feature of which lies in augmenting the "standard" model reference adaptive controller by a properly designed signal compensating for the effect of plant parameter uncertainty on the output error. One of the main performance improvement characteristics of the proposed IPAC is that the zero-state output error can be made arbitrarily small under standard model reference adaptive control (MRAC) assumptions in the case of linear systems while a similar result holds for a class of linearizable systems as well. The exponential convergence of the output and parameter errors in the presence of sufficiently rich reference inputs, remains a valid property of this controller which also achieves improved robustness in the presence of bounded disturbances and/or unmodeled dynamics as well as in the case of an adaptation switch-off.
A benchmark study of two self-organizing artificial neural network models, ART2 and DIGNET, is conducted. The architecture differences and learning procedures between these two models are compared. The performance of ...
详细信息
A benchmark study of two self-organizing artificial neural network models, ART2 and DIGNET, is conducted. The architecture differences and learning procedures between these two models are compared. The performance of ART2 and DIGNET on data clustering and signal detection problems with noise or interference is investigated by comparative simulations. It is shown that DIGNET generally has faster learning and better clustering performance on the statistical pattern recognition problems. DIGNET has a simpler architecture, and the system parameters can be analytically determined from the self-organizing process. The threshold value used in DIGNET can be specifically determined from a given lower bound on the desirable signal-to-noise ratio (SNR). The networks discussed in this paper are applied and benchmarked against clustering and signal detection problems.
In this paper, an adaptive controller with improved transient and steady-state performance using parameter mismatch compensation and high-order tuning is presented. As in previous designs, the use of a properly design...
详细信息
In this paper, an adaptive controller with improved transient and steady-state performance using parameter mismatch compensation and high-order tuning is presented. As in previous designs, the use of a properly designed signal, compensating for the effects of parameter mismatch on the output error, leads to "arbitrary" performance improvement. However, unlike previous designs based on first-order (classical) tuners, high-order tuning is necessary in order to avoid the effect of terms due to initial conditions on the output error. This constitutes a strong motivation for the use of high-order tuners in adaptive control which so far, were considered as more complicated alternatives to the classical first-order tuners. The proposed controller also provides robustness enhancement in the presence of bounded disturbances and/or a class of unmodeled dynamics or in the event of an adaptation switch-off end, as in classical model reference adaptive control, guarantees the exponential convergence of all errors to zero in the presence of sufficiently exciting reference inputs.
When actuators and sensors on a controlled mechanical structure form naturally passive pairs, which usually requires physical collocation, very robust stabilizing controllers can be constructed. If the number of such ...
详细信息
When actuators and sensors on a controlled mechanical structure form naturally passive pairs, which usually requires physical collocation, very robust stabilizing controllers can be constructed. If the number of such naturally passive pairs is small, the closed loop performance may not be satisfactory. This motivates the incorporation of additional sensors in the control design. However, the passivity property no longer holds for these sensors. Some thoughts are presented on extending the passivity based design to incorporate these non-collocated sensors by using some nominal model information to synthesize a fictitious strictly passive output. Simulation and experimental studies are presented to show the merits of this approach.< >
暂无评论