Intelligent Systems: Models and Applications

There is given an overview of generalizations of the integral inequalities for integrals based on nonadditive measures. The Hölder, Minkowski, Jensen, Chebishev and Berwald inequalities are generalized to the Choquet and Sugeno integrals. A general inequality which cover Hölder and Minkowski type inequalities is considered for the universal integral. The corresponding inequalities for important cases of the pseudo-integral and applications of these inequalities in pseudo-probability are also given.

Since interval-valued measures have applications in number of practical areas, this paper is focused on two approaches to this problem as well as on the corresponding generalizations of the Jensen and the Chebyshev integral inequalities. The first approach is based on an interval-valued measure defined by the pseudo-integral of interval-valued function, while the second approach considers an interval-valued measure obtained through pseudo-integrals of real-valued functions.

The aim of this paper is to construct a General Prioritized Fuzzy Satisfaction Problem (GPFCSP) that can handle any logical expression whose atomic symbols are prioritized constraints. GPFCSP can be applied in the field of Fuzzy Relational Databases, multilateral negotiations, decision making etc. The interpretation method is used in order to obtain a complete axiomatization.

The contribution, in its first part (chapters [1]-[4]), emphasizes some of the differences between human beings and robots, and in its second part (chapters [5]-[7]) also some differences between computers and their programming, and robots and programming of robots. It recognize programming as an integral part of the overall human culture, and formulates complementing the usual Turing hypothesis, another Turing hypothesis rooted in the famous Turing test. As an example of focusing students attention to the differences in programming computers and robots the article provides some examples of the teaching experiences of the Institute of Computer Science of the Silesian University in Opava, Czech Republic.

In this work is demonstrated the possibility of using primitives to generate complex movements that ensure motion of bipedal humanoid robots in unstructured environments. It is pointed out that for the robot’s motion in an unstructured environment an on-line generation of motion is required. Generation of motion by using primitives represents superposition of simple movements that are easily performed. Simple movements are either reflex or learned synchronous movements of several joints, and each of these movements represents one primitive. Each primitive has its parameters and constraints that are determined on the basis of the movements capable of performing by a human. A set of all primitives represents the data base from which primitives are selected and combined for the purpose of performing a complex movement.

The paper regards to benchmarking and qualitative evaluation of different autonomous quadrotor flight controllers. Three characteristic representatives of frequently used flight control techniques are considered: PID, backstepping and fuzzy. The paper aims to contribute to the objective assessment of quadrotor control performances with respect to the criteria regarding to dynamic performances, trajectory tracking precision, energy efficiency and control robustness upon stochastic internal and/or external perturbation. Qualitative evaluation of the closed-loop system performance should enable the best choice of microcopter control structure. Non-linear modeling, control and numerical simulation of two characteristic flight test-scenarios (indoor as well as outdoor) are described in the paper, too. Obtained simulation results for three representative control algorithms are graphically and table presented, analyzed and discussed.

In this paper a new mathematical model of basic planar imprecise geometric objects (fuzzy line, fuzzy triangle and fuzzy circle) are introduced. Also, basic measurement functions (distance between fuzzy point and fuzzy line, fuzzy point and fuzzy triangle, two fuzzy lines and two fuzzy triangles) as well as spatial operation (linear combination of two fuzzy points) and main spatial relations (coincidence, between and collinear)is proposed. Results obtained with our model can be used in various applications such as image analysis (imprecise feature extraction), GIS (imprecise spatial object modeling), robotics (environment models). Imprecise point objects are modeled as a union of linear combinations of fuzzy points in linear fuzzy space. However, it is proved that fuzzy line could be represented only by two and fuzzy triangle with three fuzzy points.

In the chapter basic concepts of type-2 fuzzy logic, computing with words and perceptual computing are presented. Architectural details of an object-oriented realization of a software library for developing perceptual computers are exposed and explained. Thereby, following topics are covered: mathematical models for data types, class hierarchies for types and inference operators, operation in multithreaded environment, comparison with an existing MATLAB realization, and a short code sample. Introducing a class hierarchy for inference operators is motivated by prior work of the authors that implied novel weighted averages can be replaced with other operators. Development of a perceptual computer and its usage for hierarchical decision making in solving known problem of missile system selection is described. Input values and produced results are presented in details in the chapter. The results are consistent with results from the literature. Conclusions are given, as well as possible directions for further research and application work.

This chapter gives two–degree–of–freedom (2–DOF) speed control solutions for brushless Direct Current motor (BLDC–m) drives with focus on design methodologies. A classical 2–DOF structure, 2–DOF proportional-integral (PI) and proportional–integral–derivative (PID) structures and 2–DOF fuzzy control solutions are presented and approaches regarding the methods are highlighted. A case study concerning a BLDC–m drive with variable moment of inertia is presented. Comparative studies based on digital simulation results are included to exemplify the design methods.

VDS is a system built as intelligent thinking support for assisting medical doctor in a hospital to do medical diagnosis based on the avatar of that doctor. The medical knowledge is also collected from the doctor based on his/her experience in diagnosis. The avatar construction is mimicking real doctor. The avatar interacts with patients through their voices, and other sensors to read patient mental state and physical state that are used in aligned manner to assess the patient sickness states through Bayesian network. The physical view is represented as physical ontology. The mental view is represented as mental ontology. These two ontologies aligned on medical knowledge for diagnosis and reasoning based on similarities computation. These two types of ontologies have been mapped and aligned for reasoning using a simple Bayesian Network for causal reasoning to find related query decision case based diagnosis collected from expert doctors. The system is implemented and tested. We have constructed an integrated computerized model which reflects a human diagnostician and through it; an integrated interaction between that model and the real human user (patient) is utilized for 1 ^st stage diagnosis purposes recalled as simple cases.

In this work methods are described, which are used for an individual adaption of a dialog system. Anyway, an automatic real-time capable visual user attention estimation for a face to face human machine interaction is described. Furthermore, an emotion estimation is presented, which combines a visual and an acoustic method. Both, the attention estimation and the visual emotion estimation based on Active Appearance Models (AAMs). Certainly, for the attention estimation Multilayer Perceptrons (MLPs) are used to map the Active Appearance Parameters (AAM-Parameters) onto the current head pose. Afterwards, the chronology of the head poses is classified as attention or inattention. In the visual emotion estimation the AAM-Parameter will be classified by a Support-Vector-Machine (SVM). The acoustic emotion estimation also use a SVM to classifies emotion related audio signal features into the 5 basis emotions (neutral, happy, sad, anger, surprise). Afterward, a Bayes network is used to combine the results of the visual and the acoustic estimation in the decision level. The visual attention estimation as well as the emotion estimation will be used in service robotic to allow a more natural and human like dialog. Furthermore, the human head pose is very efficient interpreted as head nodding or shaking by the use of adaptive statistical moments. Especially, the head movement of many demented people are restricted, so they often only use their eyes to look around. For that reason, this work examine a simple gaze estimation with the help of an ordinary webcam. Moreover, a full body user re-identification method is described, which allows an individual state estimation of several people for hight dynamic situations. In this work an appearance based method is described, which allows a fast people re-identification over a short time span to allow the usage of individual parameter.

This article presents a mathematical model construction approach for functioning systems. To perform functioning system analyses with the goal to establish correct work conditions, heuristical problem solving way and mathematical modeling can be used. Topological modeling is an effective tool to develop mathematical models for heterogeneous systems when the available information is insufficient. Within this article, the authors provide a theoretical background and introduce topological model elements, functions, features, and construction phases. A practical model construction process is adapted to be used for medicine tasks. A topological model for multiple diseases is developed. It is used as a mechanism to model the course of a disease and the effect of the applied therapy. Using the proposed criteria for evaluating the chosen therapy and multi-objective optimization techniques make it possible to prescribe the optimal therapy complex for an individual patient.

In the representation and processing of knowledge, the need for knowledge approximation and similarity is frequent. Concepts needed to deal with these issues are emerging; however, the unified treatment is still missing. With this in mind, we discern two sorts of knowledge, continuous represented, which we usually meet in the sensory based knowledge systems, and the other of discrete information structures, involved in automatic reasoning and processing of symbolic sequences in various contexts. The issues related to knowledge similarity and approximation are discussed in some special cases (separately following our initial knowledge division), which are spanning more general field and is used for exposition of our ideas and solution proposals. We have presented the concepts of knowledge similarity and approximation, together with appropriate scaling - degrees of similarity/approximation, in metric spaces for continuous case and the solution for discrete information spaces- DIS spaces for the later case.

Due to the benefits of container transport, containerization of goods has increased significantly in last two decades. Consequently all aspects of container transport chains should be analyzed. Formulation of mathematical models for container terminals is important for understanding and improving container terminal operations. This paper presents an overview of existing optimization problems in container terminal operations, with the goal of defining a model for quay crane operations in river container terminals. While river container terminals are important nodes in the inland transportation network, they have not yet been addressed in the literature. The efficiency of container cranes, as the lead element in the terminal, is of key importance for the performance of the container terminal. This paper presents scheduling of quay cranes in river terminals for barge reloading.

Network monitoring is a critical issue in today’s rapidly changing network environment. Existing centralized client-server based network management frameworks suffer from problems such as insufficient scalability, interoperability, reliability, and flexibility, as networks become more geographically distributed. This work describes implementation of the agent-based system for network components and services monitoring. The system is designed in a modular fashion to provide easy and efficient inclusion of diverse monitoring objects. Several types of the intelligent agents are defined to perform efficient monitoring of diverse network components, services and applications using different monitoring protocols. In proposed architecture, we use an ontology for representation of monitoring information to provide semantics for building an underlying knowledge base that not only allows agents to communicate, but also to reason with each other, enabling the desired tasks to be performed collaboratively. The system implementation is based on the XJAF (Extensible Java-based Agent Framework) agent framework and the Java EE technology.

The precise localization of pedestrians in images is a difficult problem with many practical applications in the fields of driver assistance, autonomous vehicles and visual surveillance. Localization can be treated as a subsequent step to pedestrian detection that aims at finding the exact position of pedestrians in an input image. In this work, two different approaches for pedestrian localization using neural networks with local receptive fields are presented. The first approach uses a trained ranking classifier to determine the relative order of image windows in regard to their localization quality (coverage) of the pedestrian. Localization is then performed via sampling of the window space in the vicinity of an initial detection. For the second approach, a binary classifier is trained to stepwise move an initial window towards the optimal position. Only few network evaluations are required for this method to converge, making it applicable for real-time detection systems. It is shown how the localization task can be split up into consecutive subtasks, which allows the training of a dedicated classifier for each subtask. This progressive localization scheme improves localization precision and simplifies evaluation of the resulting classifiers. Both approaches are evaluated in detail on the publicly available Daimler Pedestrian Detection Benchmark dataset and the results are compared to a standard detection approach based on non-maximum suppression.