Knowledge-Based and Intelligent Information and Engineering Systems

This paper describes the intersection search as an inference procedure for a knowledge representation scheme based on the theory of Fuzzy Petri Nets. The procedure uses the dynamical properties of the scheme. The relationships between the concepts of interest, obtained by the intersection search algorithm, are accompanied by the value of the linguistic variable expressing the assurance for the relations. An illustrative example of the intersection search procedure is provided.

The paper deals with the problem of determination of stability margin of uncertain linear discrete-time systems with uncertainty described by fuzzy numbers. Nonsymmetric triangular membership functions describing the uncertainty of coefficients of characteristic polynomial are considered. The presented solution is based on transformation of the original problem to Hurwitz stability test and generalization of Tsypkin-Polyak plot.

The aim of this paper is to simultaneously identify and estimate a non-linear autoregressive time series using a flexible neuro-fuzzy model. We provide a self organization and incremental mechanism to the adaptation process of the neuro-fuzzy model. The self organization mechanism searches for a suitable set of premises and consequents to enhance the time series estimation performance, while the incremental method selects influential lags in the model description.

Experimental results indicate that our proposal reliably identifies appropriate lags for non-linear time series. Our proposal is illustrated by simulations on both synthetic and real data.

As more and more organizations use the Service Oriented Architecture (SOA) to design and implement their information systems also the systems’ architects need the more intelligent and reliable tools. The complexity, modularity and heterogeneity of the information systems make the security evaluation process difficult. The proposed method uses multiagent approach as the most promising direction of the research. As the security evaluation requires the precise definition of the set of evaluation criteria the basic criteria for each functional layer of SOA have been presented. Also, the paper presents two algorithms where the first can be used separately for each of the particular layer of SOA and the second serves for the calculation of the generalized SOA system security level.

This paper focuses on the issue of the formal logical description of evolutions of multi-agent systems (MAS). By evolution of a MAS we mean the change of inner states of the combined MAS caused by interaction of participating agents. We introduce a general scheme of combining propositional modal languages and respective logics into a single language suitable for such descriptions. The method is based on the representation of multi-agent systems by Kripke-Hintikka models. The obtained description allows to study the question of verifiable specifications.

Deploying and maintaining software in a distributed system includes software delivery, remote installation, starting, stoping, and modifying in order to configure or re-configure a system according to user needs. This paper deals with an agent-based framework where intelligent and mobile agents provide the means to implement a distributed system and enable its evolution by taking partial or full responsibility for software deployment tasks. Agents are organised into agent teams, where one agent is the team leader responsible for planning, while the others are operational agents capable of executing a defined plan. The formal model, as well as functionality and performance issues, are elaborated. Special attention is paid to deployment strategies and their optimization, while taking into account characteristics of distributed system nodes and the network connecting them. Simulation-based evaluation of agent serialization, migration and deserialization parameters, and their influence on overall performance, is included.

Policy-based security is an effective approach to manage knowledge systems by handling all behaviors of a system thought a set of rules. This approach has such advantages as capacity to define general high-level targets, ease for configuration, and flexibility in development and maintenance. However, the resolution of conflicts is unavoidable requirement because of many elements of subjectivity as well as objectivity in administrative processes. To this end, several works have been done, and they gave concrete results. In this paper, we will propose a new approach to solve conflicts and to integrate rules in a policy. A new representation of rules is given, the distances between rules are defined as well as postulates are presented and analyzed. Algorithms for integrating policy also have been proposed and examined.

Authors are conducting research aiming to develop new interfaces that follow the mechanism of human communication, particularly focusing on human common sense. In this paper, a method is proposed which processes any ”subject” using knowledge base and an Association Mechanism. In proposed method, 27 attributes of ”subject” were judged by knowledge base. Moreover, an unknown word processing is proposed which deals with actor words which were not registered in the knowledge base. The result of the proposed method gave the correct answer in 75% of cases. If the ”not out-of-common-sense” answers were counted as part of the ”correct answers”, the correct-answer ratio rose to 96%. Therefore, if the proposed method and the existing method were combined, the correct-answer ratio was approximately 85%.

In this paper, we propose a formal logic approach to specify the system security policies and rules and their reasoning in response to queries of accessing the system resource. Especially we investigate and handle the situation where the security agent’s knowledge based on which the access decision is made is not complete. We introduce modal logic to specify and reason about a security domain, then translate the domain into an epistemic logic program [10]. We show that our approach has an expressive power to describe a variety of complex security scenarios.

Fuzzy Description Logics are logics which allow to deal with structured knowledge affected by vagueness. Although a relatively important amount of work has been carried out in the last years, fuzzy DLs are open to be extended with several features worked out in other fields. In this work, we start addressing the problem of incorporating Multi-Criteria Decision Making (MCDM) into fuzzy Description Logics and, thus, start an investigation about offering the possibility of a fuzzy ontology assisted approach to decision making.

Knowledge acquisition and capitalization to solve problems concerning artefact evolution, still called inventive design, has a certain quantity of specific characteristics. The theoretical approach we are interested in, called TRIZ (the Russian acronym for Theory for Inventive Problem Solving), when translated into a methodological procedure, can be declined into two different steps: problem formulation and problem resolution. This article presents an analysis of two of the most used knowledge bases of TRIZ during the resolution stage. These knowledge bases have been formalized by the construction of an ontology of the informal knowledge sources usually used by the TRIZ experts. This approach has permitted the design of a software architecture that eases the implementation of these bases by means of their declarative manipulation. It combines rules and description logics for populating the ontology and facilitates the access to the compiled generic knowledge that synthesizes, at an abstract level, the already encountered problems and their solutions.

In this paper we present a new methodology for Domain modeling based on Engineering Standards. We discuss some benefits of standards as guidelines for a Knowledge Based Domain modeling, potential challenges and approaches to overcome them. The benefits of using Standards as models for Domain ontologies have been shown as valid in related work and, as proof of concept, we present a case study where our methodology was successfully applied.

Nesting algorithms deal with the optimal placement of shapes in specified regions subject to specified constraints. In this paper, a complex algorithm for solving two-dimensional nesting problem is proposed. Arbitrary geometric shapes are first quantized into a binary form. These binary representations are subsequently processed by operators which nest the shapes in a rectangle of minimum area. After nesting is completed, the binary representations are converted back to the original geometric form. Investigations have shown that the nesting effect is driven by quantization accuracy. Therefore, better accuracy is possible given more computing time. However, the proposed knowledge based system can significantly reduce the time of nesting, by intelligently pairing shapes, based on prior knowledge of their form.

The paper concerns the survivability of computer network area. The considered optimization task is the congestion problem in connection-oriented networks. This paper presents a highly configurable evolutionary algorithm together with computer experimentation system supporting its efficiency analysis. The particular emphasis is placed on parameter control and tuning process of evolutionary algorithm. The results of analysis are discussed.

In the 3D animation field, the quality of productions is continuously increasing. It is a very active market with a high level of competitiveness where modest companies, in terms of budget, must reach a balance between the resources they can apply and the economic investment in a given production. Consequently, the automation of manual design processes, which are normally highly time-consuming, has become a crucial research topic for 3D animation studios. The work we are describing here presents one of these automatic tools, specifically focused on the synchronization of the speech and the lip movement of the characters, a process that is called

lipsync

. We have developed a very robust and accurate speech recognition module that together with a knowledge-based system, autonomously provides lipsync results. Additionally, the system has been integrated in the production plan of a 3D animation company, leading to drastic operator time reductions.

Prediction of travel time on road network has emerged as a crucial research issue in intelligent transportation system (ITS). Travel time prediction provides information that may allow travelers to change their routes as well as departure time. To provide accurate travel time for travelers is the key challenge in this research area. In this paper, we formulate two new methods which are based on moving average can deal with this kind of challenge. In conventional moving average approach, data may lose at the beginning and end of a series. It may sometimes generate cycles or other movements that are not present in the original data. Our proposed modified method can strongly tackle those kinds of uneven presence of extreme values. We compare the proposed methods with the existing prediction methods like Switching method [10] and NBC method [11]. It is also revealed that proposed methods can reduce error significantly in compared with other existing methods.

Linear ordering problem (LOP) is a well know NP-hard optimization problem attractive for its complexity, rich collection of testing data and variety of real world applications. It is also a popular benchmark for novel optimization and metaheuristic algorithms. In this paper, we compare the performance of genetic algorithms and differential evolution as efficient metaheuristic solvers of the LOP.

Crop rotation is a cropping system alternative that can reduce agriculture’s dependence on external inputs through internal nutrient recycling. Also, it maintains long-term productivity of lands and breaks weed and disease cycles. Decision criteria to choose among competing crop rotation systems include economic and environmental considerations. Having many cultivation parcels, selection of optimal rotation alternatives may become difficult as different issues have to be analyzed simultaneously. Thus, this work proposes to use Multiobjective Evolutionary Algorithms (MOEA) to solve a multi-objective crop rotation optimization problem considering various parcels and objectives. Three outstanding MOEAs were implemented: the Strength Pareto Evolutionary Algorithm 2, the Non-dominated Sorting Genetic Algorithm and the micro-Genetic Algorithm. These MOEAS were tested using real data and their results compared using a set of metrics. The provided results have shown to be potentially useful for decision making support.

Object recognition in digital image processing is the task of finding a particular object in an image. Although there are many pattern recognition methods developed for handling the problem of object recognition, it is still a challenging task in computer vision systems and image understanding. This paper presents a new model for object recognition using the concepts of Bayes classifier, fusion of probability measures, and the permanence of ratios.

This paper proposes a new wavelet-fractal image compression method by studying the limitation of existing wavelet based image compression methods. Initial errors occur at different levels of importance according to the frequencies of sublevel-band wavelet coefficients. Higher frequency sublevel bands would lead to larger initial errors. As a result, the sizes of sublevel blocks and super blocks would be changed according to the initial errors. The matching sizes between sublevel blocks and super blocks would be changed according to the permitted errors and compression rates.

This paper presents a tracking system for vehicles in urban traffic scenes. The task of automatic video analysis for existing CCTV infrastructure is of increasing interest due to benefits of behaviour analysis for traffic control. Based on 3D wire frame models, we use a combined detector and classifier to locate ground plane positions of vehicles. The proposed system uses a Kalman filter with variable sample time to track vehicles on the ground plane. The classification results are used in the data association of the tracker to improve consistency and for noise suppression. Quantitative and qualitative evaluation is provided using videos of the public benchmarking i-LIDS data set provided by the UK Home Office. Correctly detected tracks of 94% outperform a baseline motion tracker tested under the same conditions.

This paper presents some important aspects of cognitive informatics operated by cognitive processes in the human mind and implemented to new generation IT systems. This paper presents especially a selected class of cognitive categorization systems called UBIAS (

Understanding Based Image Analysis Systems

). The UBIAS systems are especially dedicated to support analysis of data recorded in the form of images for example medical images. Cognitive categorization systems operate by executing a particular type of human thought, cognitive and analysis processes which take place in the human mind and which ultimately lead to making an in-depth description of the analysis and interpreting reasoning process. The most important element in this analysis and reasoning process is that it occurs both in the human cognitive process and in the system’s information process that conducts the in-depth interpretation and analysis of data.

Achieving an accurate disparity map in a reasonable processing time is a real challenge in the stereovision field. For this purpose, we propose in this paper an original approach which aims to accelerate matching time while keeping a very good matching accuracy. The proposed method allows us to shift from a dense to a sparse disparity map. Firstly, we have computed scores for all pairs of pixels using a new dissimilarity function recently developed. Then, by applying a confidence measure on each pair of pixels, we keep only couples of pixels having a high confidence measure which is computed relying on a set of new local parameters.

The problem tackled in this work is the semantic interpretation and an attempt at computer automatic understanding of a 3D structure of spatially visualised coronary vessels with the use of AI graph-based linguistic formalisms. At the stage of the initial analysis, it was found that the problem is subject to numerous important limitations. These limitations result, among other things from the serious obstacles encountered in the development of a universal standard, defining the model shape of the healthy or diseased organ that could possibly undergo typical recognition. Due to this difficulties a decision was made to apply the methods of automatic image understanding for the interpretation of the images considered, which consequently leads to their semantic descriptions. For this purpose the linguistic approach was applied.

We propose an illogical discourse judgment technique using a concept association system with the aim of enabling computer-generated logical discourse. We focused on a relation of nouns and adjective phrases. Then the knowledge structure of how to use nouns and adjective phrases is modeled by arranging the relation in a point of wrongness. Also, this paper proposes a technique for detection relation of nouns and adjective phrases by creating a knowledge model from generation of response sentences. This paper discusses detecting method illogical combinations of words. We showed that this technique was able to very accurately judge illogical usages with 87% accuracy, thus demonstrating the effectiveness of the technique.

We investigate the length of event sequence giving best predictions when using a continuous HMM approach to churn prediction from sequential data. Motivated by observations that predictions based on only the few most recent events seem to be the most accurate, a non-sequential dataset is constructed from customer event histories by averaging features of the last few events. A simple K-nearest neighbor algorithm on this dataset is found to give significantly improved performance. It is quite intuitive to think that most people will react only to events in the fairly recent past. Events related to telecommunications occurring months or years ago are unlikely to have a large impact on a customer’s future behaviour, and these results bear this out. Methods that deal with sequential data also tend to be much more complex than those dealing with simple non-temporal data, giving an added benefit to expressing the recent information in a non-sequential manner.

This article presents our proposition of a methodology for knowledge acquisition based on dialectics. In fact, the central concept in dialectics is a “contradiction”, which declined according to the inventive design principles, can be considered as a set of Elements, Parameters and Values - values that need to show the opposite aspects of the contradictions. Using these approaches for knowledge acquisition permitted us to obtain very satisfying results for solving a problem of software analysis.

In this paper a method is presented which permits to automatically extract lexical-semantic relations between nouns (specifically for concrete nouns since they have a well structured taxonomy). From the definitions of the entries in a Spanish dictionary, the hypernym of an entry is extracted from the entry definition according to the basic assumption that the first noun in the definition is the entry hypernym. After obtaining the hypernym for each entry, multilayered hyponymy-hyperonymy relations are generated from a noun, which is considered the root of the domain. The domains for which this approach was tested were zoology and botany. Five levels of hyponymy-hypernymy relations were generated for each domain. For the zoology domain a total of 1,326 relations was obtained with an average percentage of correctly generated relations (precision) of 84.31% for the five levels. 91.32% of all the relations of this domain were obtained in the first three levels, and for each of these levels the precision exceeds 96%. For the botany domain a total of 1,199 relations was obtained, with an average precision of 71.31% for the five levels. 90.76% of all the relations of this domain were obtained in the first level, and for this level the precision exceeds 99%.

Visualisation is usually one of the first steps in handling any data analysis problem. Visualisations are an intuitive way to discover inconsistencies, outliers, dependencies, interesting patterns and peculiarities in the data. However, due to modern computer technology, a vast number of visualisation techniques is available nowadays. Even if only simple scatterplots, plotting pairs of variables against each other, are considered, the number of scatterplots is too large for high-dimensional data to visually inspect each scatterplot. In this paper, we propose a system architecture called AVEDA (Automatic Visual Exploratory Data Analysis) which computes a large number of visualisations, filters out those ones that might contain special patterns and shows only these interesting visualisations to the user. The filtering process for the visualisations is based on statistical tests and statistical measures.

This paper proposes a method that quantifies the similarity between documents based on the level of relevance among terms in order to deliver a search that captures the meaning of documents. More specifically, this paper proposes a method that uses a concept-base to look for relevance among different terms and calculates the degree of association between documents using the Earth Mover’s Distance. When the proposed methods were subjected to comparison tests with other methods using the NTCIR3-WEB, they achieved good results.

Mining high utility itemsets (HUIs) has been developing in recent years. However, the methods of mining from distributed databases have not mentioned yet. In this paper, we present a parallel method for mining HUIs in vertically partitioned distributed databases. We use WIT-tree structure to store local database on each site for parallel mining HUIs. The item i

th

in each SlaverSite is only sent to MasterSite if its Transaction-Weighted Utilization (TWU) satisfies minutility (minutil), and MasterSite only mines HUIs which exist at least on 2 sites. Besides, the parallel performance is also interesting because it reduces the waiting time of attended sites. Thus, the mining time is reduced more significant than that in mining from centralized database.

With the increase in the amount and complexity of information, data warehouse performance has become a constant issue, especially for decision support systems. As decisional experts are faced with the management of more complex data warehouses, a need for autonomic management capabilities is shown to help them in their work. Implementing autonomic managers over knowledge bases to manage them is a solution that we find more and more used in business intelligence environments. What we propose, as decisional system experts, is an autonomic system for analyzing and improving data warehouse cache memory allocations in a client environment. The system formalizes aspects of the knowledge involved in the process of decision making (from system hardware specifications to practices describing cache allocation) into the same knowledge base in the form of ontologies, analyzes the current performance level (such as query average response time values) and proposes new cache allocation values so that better performance is obtained.

In this paper we propose a methodology intended to improve Course Curriculum Design (CCD) tasks, using for this purpose Semantics and CBR techniques. In specific, our proposed methodology is focused on two points: (

i

) the re-use of available resources (courses, etc), and (

ii

) the application of the experience of different experts in the course creation. As a prove of concept, we present a case study where our methodology is applied for competence and course creation using the Spanish normative for vocational education domain (technical degree).

This paper proposes an automatic method using a MeSH (Medical Subject Headings) thesaurus for generating a semantic annotation of medical articles. First, our approach uses NLP (Natural Language Processing) techniques to extract the indexing terms. Second, it extracts the Mesh concepts from this set of indexing terms. Then, these concepts are weighed based on their frequencies, locations in the article and their semantic relationships according to MeSH. Next, a refinement phase is triggered in order to upgrade the frequent ontology’s concepts and determine the ones which will be integrated in the annotation. Finally, the structured result annotation is built.

In this work we explore the use of decision trees to identify the intent of a user query, based on informational, navigational, and transactional categorization. They are based on decision trees, using the C4.5 implementation. The classifier will be built from a query data set larger than any previously used, allowing the conclusions to have a greater reach. Unlike previous works, we will explore features that have not been evaluated before (e.g. PageRank) combined with features based on text and/or click-through data. The results obtained are very precise and the decision tree obtained allows us to illustrate relations among the variables used for classification determining which of these variables are more useful in the classification process.

The search for inspirational images is an important part of creative design. When identifying inspirational materials, designers search semantic domains that are different from the target domain and use semantic adjectives in combination with the traditionally used keywords. This paper describes research conducted within the TRENDS project, which aimed at developing a software tool for the needs of concept cars designers. The goal was to assist them with the process of collecting inspirational images from various sectors of influence. The paper describes the ontology tagging algorithm developed to index the images in the TRENDS database using concepts from two ontologies: a generic ontology called

OntoRo

, and a domain-specific ontology

CTA

developed for the needs of the project. The paper presents the evaluation of the developed algorithm and suggests areas for further research.

Rich media websites like Flickr or Youtube have attracted the largest user bases in the last years, this trend shows web users are particulary interested in multimedia presentation formats. On the other hand, Web Usage and Content Mining have focused mainly in text-based content. In this paper we introduce a methodology for discovering Website Keyobjects based in both Web Usage and Content Mining. Keyobjects could be any text, image or video present in a web page, that are the most appealing objects to users. The methodology was tested over the corporate site of dMapas a Chilean Geographical Information Systems service provider.

The Semantic Web (SW) originally aims at studying a system interoperability based on a shared common knowledge base (ontology). Henceforth, the SW sets its heart on a semantic coordination of community parlance representative resources (in complement to a common knowledge base shared by the users). The matter is not only to use techniques to handle a large amount of data, but also to use approaches to keep the community parlance features. Thus, Web documents and folksonomies are the main semantic vehicle. They are little structured and Natural Language Processing (NLP) methods are then beneficial to analyze language specificities with a view to automating tasks about text. This paper describes a use of NLP techniques for the SW through a document engineering application: the information retrieval in a catalogue of online medical resources. Our approach emphasizes benefits of NLP techniques to handle multi-granular terminological resources.

Although they sometimes seem harmless, hoaxes represent not-negligible threat to individuals’ awareness of real-life situations by deceiving them, and at the same time doing harm to the image of their organizations, which can lead to substantial financial losses. Spreading of hoaxes also influences the normal operating regime of networks and the efficiency of workers. In this paper we present an intelligent automatic hoax detection system based on neural networks and advanced text processing. In the developing of our system we use a database with real-life e-mail hoaxes, and an additional database with real-life e-mail messages. At the end we give brief experimental evaluation of the hoax detection system and comment the results.

A web user session, the sequence of pages a user visits at a web site, is valuable data used in many e-business applications but privacy concerns often limit their direct retrieval. A web server log file provides an approximate way of constucting user sessions without privacy concerns. It is only approximate because the same IP address as recorded in the web log often contains the requests of several concurrent users without each user being uniquely identified. Additionally, a user’s activation of the back and forward browser button is often not recorded in the web log because, in most cases, the browser retrieves the page from its own cache. We present an integer program to construct user sessions (sessionization) from web log data that includes the possible use of the back button. We present sessionization results on web log data from an academic web site and compare sessions constructed with and without the option of sessions with the back button.

In this paper, a new approach for fast information detection in DNA sequence has been presented. Our approach uses fast time delay neural networks (FTDNN). The operation of these networks relies on performing cross correlation in the frequency domain between the input data and the input weights of neural networks. It is proved mathematically and practically that the number of computation steps required for the presented FTDNNs is less than that needed by conventional time delay neural networks (CTDNNs). Simulation results using MATLAB confirm the theoretical computations.

Feature selection, the job to select features relevant to classification, is a central problem of machine learning. Inconsistency rate is known as an effective measure to evaluate consistency (relevance) of feature subsets, and INTERACT, a state-of-the-art feature selection algorithm, takes advantage of it. In this paper, we shows that inconsistency rate is not the unique measure of consistency by introducing two new consistency measures, and also, show that INTERACT has the important deficiency that it fails for particular types of probability distributions. To fix the deficiency, we propose two new algorithms, which have flexibility of taking advantage of any of the new measures as well as inconsistency rate. Furthermore, through experiments, we compare the three consistency measures, and prove effectiveness of the new algorithms.

One of the main challenges in the operation of multirobot systems is to find ways for them to adapt to changing situations and even objectives without any type of central control. In this work we propose a real time coevolutionary strategy based on Embodied Evolution (EE) approaches that provides a means to achieve this end. The main inspiration for this approach comes from the field of artificial life combined with some of the notions on the distribution of utility functions as proposed by the multiagent systems literature. The solution has been tested on different real life problems involving robot teams. In particular, in this paper the work is aimed at the coordination of sets of robots for performing monitoring and surveillance operations such as the ones required on ship tanks and hulls. Nevertheless, the approach is general enough to be applied to many other tasks in several fields.

The aim of this pioneering research is to study, design, and implement systems that could tutor other sub-systems using techniques that traditional

real-life

Teachers use when they teach

real-life

Students. The research undertaken is a result of merging the fields of Intelligent Tutoring Systems (ITS) and Learning Automata (LA), and leads to a paradigm which we refer to to as “Intelligent Tutorial-

like

” systems. In our proposed novel approach,

every

component incorporates the fundamental principles of LA. Thus, we model the Student (i.e., the learning mechanism) using an LA. Each Student is considered to be a member of a Classroom of Students, each of whom is individually represented by a distinct (and possibly different) LA. We also model the Domain and the Teacher using the LA paradigm.

Our research also works within a new philosophical perspective. We relax the constraint that “traditional” Tutorial systems have, namely the one of assuming that the Teacher is infallible. Rather, we assume that the Teacher is inherently uncertain of the domain knowledge, and is thus of a stochastic nature. However, although he is not absolutely certain about the material being taught, he is also capable of improving his

own

“teaching skills” even while the operation of the system proceeds. Finally, we also attempt to model a realistic learning framework, where the Students can learn not only from the Teacher, but also from other colleague Students in the Classroom.

This talk enumerates the challenges of building a game generator that works like the SimCity and/or empire building genre of games. This talk begins by describing a universally recurring socio-cultural “game” of inter-group competition for control of resources. It next describes efforts to author a game generator and software agents able to play the game as real humans would - which suggests the ability to study alternative ways to influence them, observe effects, and potentially understand how best to alter the outcomes of dysfunctional economies and potential conflict situations. I then examine two implemented game worlds (NonKin Village and FactionSim Countries). I conclude by arguing that substantial effort on game realism, best-of-breed social science models, and agent validation efforts is essential if analytic experiments are to effectively explore alternative ways to influence outcomes.

The advancement of service-oriented computing and mobile device technologies gives us new challenges to provide intelligent services in ubiquitous computing (ubicomp) environments. User-centricity and dynamism support are the most essential requirements to meet those challenges. In this talk, I will introduce a user-centric and intelligent service composition framework that allows users to create their personalized ubicomp applications that utilize service resources in a highly dynamic ubicomp environments. The main features of our framework include: (1) task-oriented and spontaneous service composition; (2) dynamic service monitoring and reconfiguration; and (3) pervasive service retrieval and management. I will also explain our experiences of applying this framework to urban computing applications and intelligent service robots.