Ágnes Szeghegyi

Novel Adaptive Control of Partially Modeled Dynamic Systems

The basic components of Soft Computing were almost completely developed by the sixties. In our days SC means either separate or integrated application of Neural Networks (NN) and Fuzzy Systems (FS) enhanced with high parallelism of operation and supported by several deterministic, stochastic or combined parameter-tuning methods (learning). The main advantage of using FS is evading the development of intricate analytical system models.

Non-conventional processing of noisy signal in the adaptive control of hydraulic differential servo cylinders

Hydraulic differential electric servo cylinders are strongly nonlinear, coupled multivariable electromechanical tools applicable for driving e.g. manipulators. The primary controllable physical agent in such systems is the time-derivative of the pressure of the working fluid in the appropriate chambers of the cylinder. Besides the nonlinearities of hydrodynamical origin discontinous ones described by the Stribeck model originate from the friction between the cylinder and the piston. This model contains the terms of the viscous, the static, and the adhesive contributions of friction. Whenever the velocity of the piston changes its sign alternating friction force of considerable amplitude appears. Such behavior means serious difficulty in feedback-based continuous path (CP) dynamic control whenever a nominal trajectory asymptotically approaching a zero velocity segment is needed: the system itself generates a noise-like acceleration signal to be used in the control. Since hydraulic drives have considerable advantages in comparison with electric ones it would be desirable to extend their application to dynamic CP control, too. For this purpose a special adaptive nonlinear control dealing with the system-generated noisy signals was elaborated. The method is rather based on a novel branch of soft computing slightly supported by a nonconventional noise filtering. The capabilities of the improved controller are illustrated via simulation.

Nonconventional processing of noisy signals in the adaptive control of hydraulic differential servo cylinders

Hydraulic differential electric servo cylinders are strongly nonlinear, coupled multivariable electromechanical tools applicable for driving, e.g., manipulators. Since they have considerable advantages in comparison with electric drives, it would be desirable to extend their application to dynamic continuous path (CP) control, too. From this point of view these systems have the following specialties: a) the most important phenomena influencing their behavior as, e.g., warming up of the sliding surfaces, are determined by local effects, and cannot be controlled globally; b) the friction forces show discontinuous variation at the zero transition of the pistons velocity that is a locally nonlinearizable nonlinearity. A common proportional, integral, and derivative (PID) controller may generate a noise-like acceleration signal due to feeding back the effects of such fluctuations. Warming up of the working fluid during operation influences these friction properties, too. Dynamic interaction between the system and its environment neither measured nor modeled by the controller is another agent influencing the systems observable behavior. For this purpose a special controller was elaborated. It implements certain adaptivity to compensate the effects of the inaccurate model and the unknown external disturbances, and also contains a nonconventional noise filtering technique to reduce the effect of the fluctuating friction forces. In the paper the control method is described, and then its capabilities are illustrated via simulation results.

Fixed point transformation-based adaptive optimal control using nonlinear programming

To reduce the effects of modeling imprécisions, in the traditional “Receding Horizon Control” (RHC) that works with finite horizon lengths, in the consecutive horizon-length cycles, the actually measured state variable is used as the starting point in the next cycle. In this design, within a horizon-length cycle, a cost function is minimized under a constraint that mathematically represents the dynamic properties of the system under control. In the “Nonlinear Programming” (NP) approach the state variables as well as the control signals are considered over a discrete time-resolution grid, and the solution is computed by the use of Lagranges “Reduced Gradient” (RG) method. It provides the “estimated optimal control signals” and the “estimated optimal state variables” over this grid. The controller exerts the estimated control signals but the state variables develop according to the exact dynamics of the system. In this paper an alternative approach is suggested in which, instead of exerting the estimated control signals, the estimated optimized trajectory is adaptively tracked within the given horizon. Simulation investigations are presented for a simple “Linear Time-Invariant” (LTI) model with strongly non-linear cost and terminal cost functions. It is found that the transients of the adaptive controller that appear at the boundaries of the finite-length horizons reduce the available improvement in the tracking precision. In contrast to the traditional RHC, in which decreasing horizon length improves the tracking precision, in our case some increase in the horizon length improves the precision by giving the controller more time to compensate the effects of these transients.

Fixed point transformation-based adaptive optimal control using NLP

To reduce the effects of modeling imprécisions, in the traditional “Receding Horizon Control” that works with finite horizon lengths, in the consecutive horizon-length cycles, the actually measured state variable is used as the starting point in the next cycle. In this design, within a horizon-length cycle, a cost function is minimized under a constraint that mathematically represents the dynamic properties of the system under control. In the “Nonlinear Programming” (NLP) approach the state variables as well as the control signals are considered over a discrete time-resolution grid, and the solution is computed by the use of Lagranges “Reduced Gradient” (RG) method. It provides the “estimated optimal control signals” and the “estimated optimal state variables” over this grid. The controller exerts the estimated control signals but the state variables develop according to the exact dynamics of the system. In this paper an alternative approach is suggested in which, instead of exerting the estimated control signals, the estimated optimized trajectory is adaptively tracked within the given horizon. Simulation investigations are presented for a simple “Linear Time-Invariant” (LTI) model with strongly non-linear cost and terminal cost functions. It is found that the transients of the adaptive controller that appear at the boundaries of the finite-length horizons reduce the available improvement in the tracking precision. In contrast to the traditional RHC, in which decreasing horizon length improves the tracking precision, in our case some increase in the horizon length improves the precision by giving the controller more time to compensate the effects of these transients.

Adaptive control of a dynamic system having unmodeled and unconstrained internal degree of freedom

A new branch of computational cybernetics based on principles akin to that of the traditional soft computing (SC) was recently developed for the control of inaccurately modeled dynamic systems under external disturbances. In the present paper the operation of this controller is studied in the case of an incompletely modeled dynamic system, that is when the system to be controlled contains internal degree of freedom not modeled by the controller. As starting point the method uses a simple, incomplete dynamic model to predict the propagation of the state of the modeled degrees of freedom also influenced by that of the unmodeled internal ones by nonlinear coupling. The controller is restricted to the observation of the behavior of the generalized coordinates the models of which are available for it. By the use of a priori known, uniform, lucid structure of reduced size, simple and short explicit algebraic procedures especially fit to real-time applications the controller is able to learn the behavior of the observed system. Simulation examples are presented for the control of a double pendulum-cart system in which the first pendulum and the linear degree of freedom of the cart has drives only. The second pendulum can move freely and serves as the unmodeled component. Rotation of the second pendulum influences the inertia matrix of the whole system. It can obtain potential energy via the inertial and gravitational forces. It is found that the adaptive controller can successfully cope with the problem of imperfect modeling.

Generation of manufacturing process model entities

The application of advanced geometric and technical models of mechanical products to the planning of manufacturing processes and CNC programs is one of the most widely investigated topics. Utilization of the flexibility of a flexible manufacturing system for fulfilling customer demands is impossible without flexible planning of manufacturing processes. This paper presents an approach to generic modeling of manufacturing processes and a method for creating Petri net representations of manufacturing process model entities. Manufacturing task features and process variants are taken into account. The Petri net process model is evaluated in the course of process planning and production scheduling. Process models are created on the basis of a form feature oriented part model and a workshop environment model. Petri net objects are created using knowledge on manufacturing process objects and process structure. The paper is organized as follows. First, the aims of the research are outlined and the characteristics of manufacturing process models of this type are summarized. Next, the principle of modeling, the structure of the process model and the process model entities are detailed. Following this, the generation procedure of the process model is presented. Finally, typical examples of manufacturing process model entities for a prismatic part machined in a flexible manufacturing system are shown.

Calibration of Leontief Input/Output table parameters

Leontief Input/Output tables are widely used for solving various economic problems. This algorithm predicts the effect of changes in an economy sector on others. The same model can also be used for solving traffic forecasting problems. The solution algorithm applied is the so-called RAS one where a certain exponential opposition function is used. A simple and fast algorithm for the determination of the parameters of this opposition function will be presented.

Improvement of fuzzy logic robot controllers using inverse entropy based T-operations

In this paper a theoretical approach to performance improvement of fuzzy logic robot controllers is presented. The approach is based on two new sets of T-operations introduced by the authors. The T-norms and T-conorms are defined as minimum and maximum entropy operations. Simulation has been carried out so as to compare the effects of the new and the conventional T-operators in case of a 4 DOF rigid-link flexible-joint SCARA type robot. It is concluded that by fixing the other parameters of the controller certain sets of T-operations can improve the performance of the controller.

Fuzzy logic controllers using generalized operators defined on the basis of certainty function

In this paper new generalized operators, based on a fuzzy entropy approach are presented. After a general discussion on fuzzy entropy the concept of elementary certainty function of a fuzzy set is introduced. Using this mapping, the generalized intersections and unions are defined as mappings, that assign the less and the more certain membership grade to each of the elements of the domain of the operators, respectively. It is shown that these operators satisfy the axiom system of generalized operations and can be constructed from the conventional min and max operations. The second part of the paper investigates the applicability of the new operators in fuzzy logic controllers. Simulations have been carried out so as to determine the effects of the operators on the performance of the fuzzy controllers.