Partially observable Markov decision process (POMDP) is a useful technique for building intelligent tutoring systems (ITSs). It enables an ITS to choose optimal tutoring actions when uncertainty exists. An obstacle to applying POMDP to ITSs is the great computational complexity in decision making. The technique of policy trees may improve the efficiency. However, the number of policy trees is normally exponential, and the cost for evaluating a tree is also exponential. The technique is still too expensive when applied to a practical problem. In our research, we develop a new technique of policy trees for better efficiency. The technique is aimed at minimizing the number of policy trees to evaluate in making a decision, and reducing the costs for evaluating individual trees. The technique is based on pedagogical orders of the contents in the instructional subject. In this paper, we first provide the background of ITS and POMDP, then describe the architecture of our POMDP based ITS, and then present our technique of policy trees for POMDP solving, and finally discuss some experimental results.

When an intelligent tutoring system (ITS) teaches its human student on a turn-by-turn base, the teaching can be modelled by a Markov decision process (MDP), in which the agent chooses an action, for example, an answer to a student question, depending on the state it is in. Since states may not be completely observable in a teaching process, partially observable Markov decision process (POMDP) may offer a better technique for building ITSs. In our research, we create a POMDP framework for ITSs. In the framework, the agent chooses answers to student questions based on belief states when it is uncertain about the states. In this paper, we present the definition of physical states, reduction of a possibly exponential state space into a manageable size, modelling of a teaching strategy by an agent policy, application of the policy tree method for solving a POMDP, and online teaching strategy improvement. We also describe an experimental system implementing the framework, some initial experimental results, and result analysis.

Detecting and monitoring odour around livestock farms are difficult. In this paper, a dynamic neural network based model is proposed to locate odour dispersion around livestock facilities. The proposed dispersion model can dynamically represent complex or non-steady-state meteorological and topographical features in and around livestock farm areas. The proposed approach can also model odour dispersions from multiple odour sources, and from various types of sources such like point source, line source, and area source. In addition, the proposed model simulates the odour dispersion through the dynamic neural activity landscape, without explicitly additional models of the dynamic environment, odour sources, and farming activities.

Real-time fine motion control of a nonholonomic mobile robot is investigated, where both the robot dynamics and geometric parameters are completely unknown. A neural network controller combining both kinematic control and dynamic control is developed. The neural network assumes a single layer structure, by taking advantage of the robot regressor dynamics that express the highly nonlinear robot dynamics in a linear form in terms of the known and unknown robot parameters. The learning algorithm is computationally efficient. The system stability and the convergence of tracking errors to zero are rigorously proved using a Lyapunov stability theory. The real-time fine control of a mobile robot is achieved through the online learning of the neural network. In addition, the developed controller is capable of learning the kinematic parameters online. The effectiveness and efficiency of the proposed controller is demonstrated by simulation studies.

Polygon boundaries on thematic maps are conventionally considered to be sharp lines representing abrupt changes of phenomena. However, in reality changes of environmental phenomena may also be partial or gradual. Indiscriminate use of sharp lines to represent different types of change creates a problem of boundary inaccuracy. Specifically, in the context of vector-based GIS, use of sharp lines to represent gradual or partial changes may cause misunderstanding of geographical information and reduce analysis accuracy. In this paper, the expressive inadequacy of the conventional vector boundary representation is examined. A more informative technique—the fuzzy representation of geographical boundaries—is proposed, in which boundaries describe not only the location but also the rate of change of environmental phenomena. Four methods of determining fuzzy boundary membership grades from different kinds of geographical data are described. An example of applying the fuzzy boundary technique to data analysis is presented and the advantages of the technique are discussed.

. A fuzzy query approach is presented which is aimed at introducing more natural language expressions into GIS user interfaces. In this approach “vague”natural language expressions are represented as fuzzy formulae and are processed in a systematic manner. This approach enables modification-type, composition-type and quantification-type fuzzy formulae to be used for data retrieval. This approach can be easily implemented, including on most of the existing nonfuzzy systems. The ease of implementation makes this approach attractive for developing new systems and for enhancing existing GIS. In this paper, the concepts underlying the approach and the methods for processing fuzzy formulae are described in detail. A prototype for implementing this approach and a group of sample queries are described.

An experimental vision system for remote sensing land use/cover change detection is presented. The system was designed to monitor the processes of urban development. The design strategy, system structure, and image analysis sequence are addressed. Some experimental results are presented. Techniques are described in detail which were aimed at addressing the key aspects to constructing a knowledge-based vision system: preprocessing, spatial representation and manipulation, and incorporating ancillary data.>

The major barriers to the integration of remote sensing image analysis and geographical information system (GIS) technologies include the problems with data format, data accuracy, and system communication. This research explores a new path for achieving the integration. The three problems, which at first may seem to be independent of each other, are considered as different aspects of a problem with knowledge (information) representation and are tackled by unifying the systems on a common knowledge representation framework. Three new techniques-relational-linear quadtree, fuzzy representation method, and communication subschema-have been developed to provide such a framework.>

An approach called relational-linear quadtree which provides information at multiple resolutions to describe spatial features together with descriptive or propositional information in desired structures is presented. It also provides powerful and flexible spatial analysis and manipulation ability. Its principle is quite clear, and implementation is very easy. It can be modified to fit other database models or structured knowledge representation schemes and can be implemented on most database management systems (DBMS) or expert system tools which support the structured models or schemes. The only extension is a set of quadtree operators which involve arithmetic operations on integers that can be coded as routines in a host language or a system command language. No additional processor is needed. This advantage makes the relational-linear quadtree a very practical approach in making use of the existing DBMSs or expert system tools to develop spatial information processing systems.>

Currently used methods for representing geographical information are inadequate because they do not tolerate imprecision. This leads to information loss and inaccuracy in analysis. Such expressive inadequacy is largely due to the underlying membership concept of classical set theory. To improve information processing in GIS research and application, an alternative membership concept is required. In this paper, we explore the inadequacy imposed upon geographical information representation by classical set theory and address the problems of information loss. A fuzzy relational data model is defined which is more representative for geographical information. A GIS database for agricultural land resource management is created by using the data model and a new technique for assessing land suitability is developed. The fuzzy representation largely facilitates data analysis in this GIS. The methods are tested with data from North Java, Indonesia using a vector-based GIS software package, Arc Info, and the analysis results are presented.