Showing papers in "Intelligent Automation and Soft Computing in 1996"

Generalization of Rough Sets using Modal Logics

Abstract: The theory of rough sets is an extension of set theory with two additional unary set-theoretic operators defined based on a binary relation on the universe. These two operators are related to the modal operators in modal logics. By exploring the relationship between rough sets and modal logics, this paper proposes and examines a number of extended rough set models. By the properties satisfied by a binary relation, such as serial, reflexive, symmetric, transitive, and Euclidean, various classes of algebraic rough set models can be derived. They correspond to different modal logic systems. With respect to graded and probabilistic modal logics, graded and probabilistic rough set models are also discussed.

On Genetic Programming of Fuzzy Rule-Based Systems for Intelligent Control

Abstract: Fuzzy logic and evolutionary computation have proven to be convenient tools for handling real-world uncertainty and designing control systems, respectively. An approach is presented that combines attributes of these paradigms for the purpose of developing intelligent control systems. The potential of the genetic programming paradigm (GP) for learning rules for use in fuzzy logic controllers (FLCs) is evaluated by focussing on the problem of discovering a controller for mobile robot path tracking. Performance results of incomplete rule-bases compare favorably to those of a complete FLC designed by the usual trial-and-error approach. A constrained syntactic representation supported by structure-preserving genetic operators is also introduced.

Rough sets: Present state and further prospects

Abstract: W: Institute of Computer Science Report 49/94. Warsaw University of Technology, Poland, 00-665 Warsaw, Nowowiejska 15/19, 1994

Adaptive Decision-Making by Systems of Cooperating Intelligent Agents Organized on Rough Mereological Principles

Abstract: We propose a new approach to tasks of Distributed Artificial Intelligence (DAI). This approach is based on a novel idea of rough mereology which offers a framework for a rigorous (numerical) treatment of relations of being a part in a degree and allows for approximate reasoning about complex objects in particular for organizing systems of intelligent agents into schemes (assembling teams) for the purpose of synthesis from (elementary) parts of complex objects conforming to a given (possibly incomplete) specification in satisfactory degree.The above topics are illustrated by an example. Our example is a tiny fragment of a real animation system constructing subsequent scenes (or movie takes) from a database of elementary pictures. We expect that our general approach can be applied to practical tasks like manufacturing, design, control, management, etc.

Speaker-Independent Recognition of Digits—Experiments with Neural Networks, Fuzzy Logic and Rough Sets

Abstract: The aim of the presented research was to elaborate and to test the speaker-independent system for the man-machine voice interfacing using a small vocabulary containing digits. The trajectory tracking method was implemented to feature vector extraction from speech signal. Some intelligent decision systems were tried as the neural network algorithm supported by the fuzzy logic classification procedure and the rough set method permitting to derive decision rules for the recognition of speech patterns. Implemented methods will be described. Results of comparative studies in the domain of speaker-independent recognition of digits will be presented. Conclusions on the effectiveness of implemented decision systems will be added.

Cooperative Knowledge-Based Systems

Abstract: Traditional query processing provides exact answers to queries. It usually requires that users fully understand the database structure and content to issue a query. Due to the complexity of database applications, incorrect or incompletely specified queries are frequently posed and the users often receive no answers or they might need more information than they have received. In this paper a multi-agent system called a cooperative knowledge-based system (CKBS) is presented to rectify these problems.

Rough Set and Fuzzy Set Methods Applied to Acoustical Analyses

Abstract: A novel approach to computer assessment of acoustical quality has been made using soft computing methods. Rough set and fuzzy set theories proved to be especially interesting in applications to acoustical assessments. Objective measurement and listening test results were collected in order to provide data for the rough set computations. Correspondingly, examples of automatic assessment of sound quality were obtained. Moreover, some exemplary data derived from performed subjective tests were processed using a fuzzy set-based method. Conclusions concerning the artificial intelligence approach to processing of acoustic data were included.

PD-Type Control Schemes for Cooperative Manipulator Systems

Abstract: Kinematic control of cooperative manipulator systems is based on the solution of an inverse kinematics problem providing the single-arm reference joint trajectories corresponding to the assigned cooperation task. If an independent joint control scheme is used to realize the desired joint trajectories, tracking error components causing violation of closed-chain constraints lead to building of large internal forces. A simple PD-type control law is presented in this article which is based on kineto-static filtering of the joint error components causing internal forces. The control law retains the simplicity of independent joint control schemes and achieves null internal force at steady state. The addition of internal-force feedback is also analyzed. A case study for a planar cooperative system is developed which shows the effectiveness of the proposed kineto-static filtering and makes a comparison with classical PD control.

Path Planning for Mobile Robots via Skeletons on Fuzzy Maps

Abstract: In this article, we present algorithms for robot path planning via navigation functions on fuzzy maps. In these kinds of bitmaps, a number is associated to each cell, characterizing the possibility that it belongs to an obstacle. Such a representation is useful when unknown environments are reconstructed from sensor measures. Two methods are devised; both are based on a proper modification of the wavefront expansion algorithm. The first is relatively simple, but produces paths that may graze the obstacles. Instead, the second method makes use of a properly defined skeleton of the map, that lies as far as possible from such areas. The performance of the two methods is illustrated by application to both simulated and experimental maps. The proposed method is also extended to the case of nonholonomic wheeled mobile robots.

On Discovery of Attribute Interactions and Domain Classifications

Abstract: This article presents the main ideas of an approach to database mining involving search for complete, or close to complete, domain classifications in terms of attribute values. The objective is to find classifications which are based on interacting or related attributes and provide good characterization of the concept of interest by maximizing predefined quality criteria. This article introduces the notion of the classification complexity and several other measures to evaluate connections among attributes, classifications, and the concept of interest. In addition to that, an algorithm for searching for classifications is also presented. The approach is illustrated using results of a commercial market database analysis.

Dynamic Allocation of Buffer Capacity in Discrete Event Systems

Abstract: In this article, the problem of dynamic allocation of buffer capacity in a discrete event system is studied, and a solution is proposed. This problem is of interest, among others, in Kanban manufacturing systems, where the buffer capacities are equal to the number of Kanbans in each working centre.The contribution of the article consists of an on-line control strategy, using estimates of the system performance for the dynamic allocation of buffer capacity in order to increase the performance of the system, despite unknown disturbances entering the system. The system performance estimation procedure is based on a simple adaptation of already known finite perturbation analysis rules. The proposed control strategy has been tested by means of simulation on a multi part-type Kanban system. In this case, the control action specializes in the dynamic allocation of Kanbans to each part-type in order to counteract the effects of unknown changes in the product-mix on the system throughput. Several simulatio...

Training Asymptotically Stable Recurrent Neural Networks

Abstract: In this work we present a class of recurrent networks which are asymptotically stable. For these networks, we discuss their similarity with certain structures in the central nervous system, and prove that if an interconnection pattern that does not allow excitatory feedback is used, then the resulting recurrent neural network is stable. We introduce a training methodology for networks belonging to this class, and use it to train networks that successfully identify a number nonlinear systems.

Fuzzy Speed Regulation of A Wind-Turbine DC Generator

Abstract: A blade-pitch angle control technique based on fuzzy-logic formulation is presented in this article. The technique takes advantage of fuzzy-linguistic modeling in expressing the natural non-linearity or imprecision of the wind-turbine system model in designing the pitch control for speed/power regulation. Performance of the pitch control design method is demonstrated on a simulated autonomous DC generating unit connected to a variable load.

Modeling the Response of An Electrorheological Fluid Damper: Constitutive Models and Neural Networks

Abstract: This article presents three models for predicting the response on an electrorheological (ER) fluid damper developed for seismic protection of structures. The proposed ER damper is a semi-active device and the resulting damping forces can be controlled by applying an electric field. First the performance of a simple nonlinear phenomenological model is investigated. Subsequently, a neural network model is evaluated. Finally, a combination of the phenomenological model and the neural network is presented to approximate the response of the ER fluid damper. The performance of the proposed models is validated against experimental data.

Machine Discovery of Functional Components of Proteins from Amino-Acid Sequences Based on Rough Sets and Change of Representation

Abstract: Protein structure analysis from DNA sequences is an important and fast growing area in both computer science and biochemistry. Although interesting approaches have been studied, it is very difficult to capture the characteristics of protein, since even a simple protein is made of more than 100 amino acids, which makes biochemical experiments very difficult to detect functional components. For this reason, almost all the problems in this field are left unsolved and it is very important to develop a system which assists researchers on molecular biology to remove the difficulties caused by combinatorial explosions. In this article we report a system, called MW1 (Molecular biologists' Workbench version 1.0), which extracts knowledge from amino- acid sequences by controlling the application of domain knowledge automatically. We apply this method to comparative analysis of lysozyme and α-lactalbumin. The results show that we obtain several interesting results from amino-acid sequences, which have not be...

A Sonar Based Obstacle Avoidance Module for Application On AUVs

Abstract: This article concerns the development of a Collision Avoidance Module (CAM) to be integrated into a sonar-based navigation and mapping System for guidance and control of autonomous underwater vehicles. The CAM employs a Kalman filter, for dealing with the presence of uncertainty in the sensory data, and it can work at different levels of data abstraction, in order to cope with different situations. These features enable the CAM to generate avoidance strategies without making use of a priori information about the possible obstacles or their motion. The performances of the CAM are evaluated and discussed using simulation.

On Using Recurrent Neural Network Models To Develop Direct Robust Adaptive Regulators For Unknown Systems

Abstract: A direct nonlinear adaptive controller, to solve the regulation problem for unknown dynamical systems that are modeled by recurrent neural networks is discussed. The behavior of the closed-loop system is analyzed for the case in which the true system differs from the recurrent neural network due to the presence of a modeling error term. Generally, our adaptive regulator can guarantee uniform ultimate boundedness of the state and boundedness of all signals in the closed loop. Furthermore, the magnitude of the growth of the modeling error is considered unknown.

Determination of Contact Forces for Two Manipulators Mutually Lifting a Rigid Object Using A Technique of Dynamic Load Distribution

Abstract: This article presents a new technique and approach for dynamically distributing the payload when two serial link manipulators mutually hold and transport a rigid object. A new vector variable which parameterizes the internal stress contact force and torque degrees of freedom (DOF) is introduced and defined as a linear function of the contact forces.” When this equation is combined with the object's dynamical equations, the resulting system yields a well specified solution for the contact forces as a function of the internal stress DOF parameterizing variable. The article investigates two different approaches to selecting the matrix which transforms the contact forces to obtain the internal force parameterizing variable. In one approach the matrix is selected to be orthogonal to the transformation matrix in the object's dynamical equations and in the other it is a function of two constrained scalar parameters which govern the distribution of the payload between the object motion inducing components...

Robust Learning Control by Neural Network for an Active Four Wheel Steering System

Abstract: The dynamics of a four wheel steering(4WS) system inherently has model uncertainties, resulting in degradation of vehicle handling performance. To compensate for model uncertainties of the vehicle system, a nonlinear neural network control scheme is proposed and evaluated. The control scheme is composed of a conventional model reference control term and a compensator term. The compensator term is generated by a neural network whose teaching signal is the error between the actual plant and the reference model. This control scheme does not require an inverse dynamics of the plant or a Jacobian information of the learned plant in order to carry out on-line learning. Since the teaching signal of this scheme is simple to compute, fast convergence can be realized. Adaptive capability of the neural network compensator for the structured uncertainties has been demonstrated. The validity and effectiveness of the proposed control scheme for a vehicle four wheel steering are verified by computer simulations....

Identification of Robot Kinematic Parameters Using Evolutionary Algorithms

Abstract: This article proposes an alternative heuristic solution based on Genetic Algorithms (GAs) to the manipulator kinematic parameters identification problem. GAs, which refer to a family of algorithms that rely on analogies to natural evolution, have been applied to a wide range of optimization problems with extremely encouraging results. GAs outperform classical optimization techniques since they do not suffer from the disadvantages that are inherent to these methods. The formulation and the application of Genetic Algorithms for the solution of the above problem are discussed and the concept is verified by simulations in which the kinematic parameters of serial and parallel manipulators are identified. The results obtained demonstrate the simplicity and efficiency of this approach.

Evidential Comparisons Using Belief Functions, Rough Sets and Nonmonotonic Preferences

Abstract: This study establishes a relationship between the evidential comparisons in the rough set theory, the nonmonotonic preference structures, and the theory of belief functions. The nonmonotonic preference structures use the notion of existing as well as future evidence. Hence, these structures can be used to describe the accumulation of evidence in belief functions, and the knowledge refinement in the rough set theory. The consistency between combination operations in these three theories may be useful for the development of a unified decision theoretic framework.

Constructing Abstract Belief Networks that Preserve Probabilistic Orderings: An Algebraic System Approach

Abstract: This article presents an algebraic approach for abstracting belief networks. The goal of abstracting belief networks is to discover alternative belief networks with simple network structures that preserve probabilistic orderings. Since the topological structure of a belief network dictates inferential complexity, abstracting belief networks is an alternative for improving the efficiency of probabilistic inference.

A Virtual Manufacturing Workcell for Automated Assembly

Abstract: This work describes the implementation of a novel robot workcell programming interface that allows an assembly designer to obtain immediate feedback regarding the manufacturability of his/her design. The interface allows the user to manipulate the three-dimensional CAD/CAM models of the components and “assemble” them into the final product. The computer then analyzes the relevant assembly operations and translates them into low-level commands for the robots in the specific workcell under consideration. This work is motivated by the complexity and time-consuming nature of manually programming flexible assembly cells for the manufacture of different products, particularly when they involve the cooperation of multiple robot manipulators.

Quasi-Static Modeling of Compliance-Governed Redundant Manipulators

Abstract: A compliance-governed (redundant) manipulator is the main subject of the article, and its global quasi-static behaviors are studied. To implement those behaviors in an actual manipulator, mathematical compliance paradigms in three forms are proposed. They are: kinematic (or equilibrating motion) level, quasi-static (or transient motion) level, and stiffness level compliance paradigm. Each has its own usefulness and implication as described in this article. Repeatable kinematic behavior of a compliance-governed manipulator can be implanted into the inverse kinematic algorithm for redundant manipulator either by the kinematic level or quasi-static level compliance paradigm. If the transient behavior with stiffness characteristic is required in a non-compliance-governed manipulator, we will use the stiffness level compliance paradigm. It makes the arm respond equivalently to the intended compliance-governed manipulator.

A Neural Network Based Technique for Establishing Priority In CSMA/CD Scheduling

Abstract: Today, the most common local area network implementations are based on the CSMA/CD scheduling scheme. The commercial versions of the CSMA/CD based local area networks are known as Ethernets. The Ethernets provide fair, non-prioritized access to all stations on the network at all traffic loads, use a highly reliable bus topology, provide random access for all stations and employ a simple conflict resolution method of scheduling using the exponential truncated binary backoff method. In a CSMA/CD scheme, when a station needs to transmit, it simply starts sending packets of data. If no conflict is detected, transmission is assumed to be completed. In case of a conflict, all stations in the conflict, independently, calculate the amount of backoff time for scheduling retransmission. Normally, in data communications environments where some level of priority treatment of data or stations is needed, CSMA/CD based networks cannot be used. Therefore, a generalized approach to provide priority in CSMA/CD base...

Stable Adaptive Control of A Class of Nonlinear Dynamic Systems Using RBF Networks

Abstract: Based on Radial Basis Function neural networks, a stable adaptive control scheme is proposed for a class of unknown nonlinear dynamic systems. It is shown that the scheme is stable and convergent in the sense that all the signal in the loop remain bounded, and in the ideal case, output tracking error will converge to zero and in the non-ideal case, the error will converge to a residue which is proportional to the modelling error bound. Simulation results are also provided to demonstrate performance of the scheme.

Efficient Trajectory Planning and Compliant Control for Two Manipulators to Deform Flexible Materials

Abstract: Coordinating two robot manipulators to handle flexible materials has a wide range of applications in the manufacturing industry. However, this problem has not been seriously addressed until recently. The two robot manipulators have to follow complicated trajectories to maintain a minimum interaction force with the flexible beam. These trajectories are very complicated and not suitable for real time systems. In this article, three approximation methods of the optimal trajectories and a compliant control scheme are introduced. The first one uses a piece-wise linear approximation of the optimal trajectories, while the second one uses adaptive piece-wise linear approximation. The third method applies a continuous approximation of the optimal trajectories using an ellipsoid. Finally, a compliant motion scheme is proposed to reduce the interaction forces and moments in the first method. The stability of the proposed system is investigated. Experimental results encourage the proposed schemes.

An Algorithm for Data Reduction in Learning from Examples

Abstract: The input data for machine learning from examples are usually presented in a decision table. In such a table, examples are described by values of variables: attributes and a decision. A red...

A Control System for Flexible Manufacturing

Abstract: The application of computers to the manufacturing industry has substantially altered die control procedures used to program a whole manufacturing process. At present one of the major problems with Flexible Manufacturing Systems is the lack of a good practical overall control system. This paper identifies the problems involved in current methods of control and their advantages and disadvantages. As a result, a different type of control system is proposed which has been successfully developed and implemented. This control model combines the advantages of Hierarchical and Hierarchical control systems and utilizes distributed decision making and control features. The hybrid control system is comprised of four modules; manufacturing database, scheduler, communication and cell controller builder modules.

On Biased Measures and Typical Values

Abstract: Given an arbitrary fuzzy set, there are various linguistic expressions which represent different approaches to the concept of typicality and may often be transformed to suitable numerals—typical values. These quantities can be then replaced (not uniquely) by integral forms; namely, the Riemann-Stieltjes integral or the fuzzy integral. A more natural approach for obtaining typical values is to apply an inverse procedure: Based on the linguistic expression and the fuzzy set, a biased measure is designed such that the Riemann-Stieltjes integral and the fuzzy integral over the fuzzy set with respect to this measure, yield the typical value and the balance point of the set respectively