Recently, computer-assisted instruction, CAI for short, attracts a considerable attention as the computer technology develops. The goal of CAI research is to realize a high performance education with the aid of a computer. The state of the art of the current CAI systems, however, is far from the goal. Thus, a certain breakthrough is badly needed to overcome this difficulty. One candidate is artificial intelligence (AI) techniques which is becoming popular these days. A CAI system utilizing AI techniques is called an ICAI system or an Intelligent Tutoring System (ITS for short). This paper is concerned with a brief overview of the current state of the art of ITS research. Section 2 deals with what an ITS is where an ITS is defined as a system which realizes an adaptive and bi-directional interaction aiming at effective knowledge communication. By adaptive, I mean the system can give instruction appropriate to the individual student through dynamic control dependent on the the learning process of the student. And by bi-directional interaction, I mean a mixed-initiative interaction in which the student can ask questions or ask the system to solve problems. Section 3 discussed the problem of student modeling methodologies proposed to date. Several ITS's developed in Japan, BOOK, GEOMEX, a general framework, CAFEKS, SBM, GTS and INTELLITUTOR are described in Sect. 4. Furthemore, as typical ITS's developed in US, SOPHIE, STEAMER, QUEST, MENO, GUIDON are also described in Sec. 5. Future works are discussed in concluding remarks.

This paper proposes a method for producing a soft scenario" that coordinates highly individualized and mixed-initiative tutoring with strategic dialogue composed of semantically consecutive topics. The soft scenario is generated in three steps. The first step generates an outline plot of the scenario at the beginning of learning. The outline is composed of parts of domain knowledge to be taught and their sequence. A knowledge representation model, called MHM, and a method to represent author's teaching strategy, called a strategy graph, are introduced and used for generating the outline plot. The second step generates detailed parts of the scenario at the time to punctuate learning. Detailed planning is composed of two portions. One portion is to modify the pre-generated outline plot if the results of learning do not agree with the forecast. The other portion decides a teaching paradigm and a topic. A method of identifying student error origins, a teaching paradigm called adaptive guiding and information about leaning situation are presented and used for detailed planning of soft scenario together with a resolution tree of a given topic and compatibility of a topic to paradigm. The third step, which is not included in this paper, creates a dialogue part of the scenario according to a dialoguing strategy by using the previous context, and accords natural language outputs with graphical outputs.

This paper presents a script-based monitoring system for evaluating students' reasoning during problem-solving. Problem-solving processes are equipped with knowledge structures, which are named scripts. A script is a collection of slots for events and provides conceptual dependency between events. Therefore, Script-Based Monitoring Systems (SBMS) can issue assistance to students' requirements and explanations to incorrect activities immediately during problem-solving. The SBMS works as an assistant for circuit analysis. A solution script is designed for the problem-solving process of sinusoidal steady-state network problems by applying loop or branch current analysis. The script is broken up into three scenes: defining variables, deriving equations, and solving equations. After a problem is displayed to a student, she explains her own problem-solving process by sentences and equations. The purpose of the SBMS is to evaluate the student input and determine the control of the dialogue at every step. The evaluation of such events is done by the diagnostic module of the SBMS and the slots of the solution script are filled with the results. As a consequence, the SBMS can hold mixed-initiative dialogues with students. The SBMS is implemented in PROLOG.

This paper is about describing the architecture of an Intelligent Tutoring System which can diagnose and give individual advice for the purpose of teaching the operational procedures and theorem-understanding in set-operations. From the technological point of artificial intelligence, a method of processing of mathematical equation for set-operational procedures and a dynamic construction of a student model based on condition induction are proposed. That is the architecture of ITS on teaching the algorithm for the operational procedures and modelling the structure of student's understanding in set-operations. The fundamental concepts consist of two kinds within the student model which are the state-memory of target achieving (MTGA) and the state-memory of problem-application (MPA). The system works cooperatively and adaptively monitoring MPA and MTGA, in order to realize more individualized tutoring corresponding with the state of the student's understanding. As the global teaching strategy, there is the selection-function of the problem matched with the level of a student's understanding pased on the method of condition-induction.

Osaka Gas has been investigating artificial intelligence for 6 years, focusing especially on research into ICAI. We have developed a simulation based GTS (Generic Training System) in cooperation with Vanderbilt University, in Tennessee USA. This system is one of the few practical ICAI systems used for real fields, and the coursewares can be developed without any programming skill. Topics covered in this paper include the system architecture, system modules and main features. As an example, a model-based intelligent tutoring system for power distribution training systems (PDTS) is also provided. The system is in use at several places in Osaka Gas.

We propose a knowledge-based instruction system for writing beautiful Japanese characters correctly, by using ICAI techniques. With this kind of CAI, it is important to relate the character pattern with the skill knowledge of handwriting because the rules of writing are indirectly shown as a feature of the character pattern. However, the CAI for penmanship previously reported was only based on the pattern analysis of the character. In this paper, we describe the rules of writing for this system as the skill knowledge of penmanship in the declarative way. This system evaluates and reforms the student's handwriting based on the rules of writing stored in a knowledge base. We would also introduce the error knowledge by the student's frequent handwriting mistakes. Finally, this system can estimate the student's degree of handwriting ability by their input charecters and give suitable guidance for improving their skills of writing. Students can learn the skills and the knowledge of penmanship effectively by this system. This system is also useful for foreign students learning the Japanese language.

An all digital 256 QAM modulator including the carrier signal of nearly 10 MHz, was constructed. In order to achieve, the very high-speed operation of nearly 100 MHz is required. In this experiment, the very high-speed digital devices such as digital multiplier with delay time of 4.3 ns was employed. Excellent modulation performance was achieved with no adjustment. Roll-off spectal shaping filter using a new BTF circuits was constructed. The excellent transmission performance was obtained and the transmission capacity of nearly 20 Mbit/s was achieved. The results provide prospects for an all digital LSI-based high-speed 256 QAM modem for high-capacity Digital Microwave Radio.

This letter illustrates the characteristics and causes of the noise voltage induced in adjacent circuit having resistive terminations at the break of silver contacts with inductive load.

The authors' developed algorithm of position and intensity estimation of reflection sources is examined to evaluate its effectiveness for practical use. The algorithm is applied to analyse the reflected waves in an ordinary room based on an EM field pattern measured by using a robotic system. The result shows that the algorithm is useful to analyse the indoor EM reflection.

This letter is concerned with application of a spectral domain method combined with the generalized sampling theorem to electromagnetic problems. Discretizations are performed by selecting unequally-spaced sampling points coincident with the zeros of an entire function of exponetial type. Numerical examples are also given for the scattering by a plane metallic grating, and it is demonstrated that the numerical results show a good convergence.

A parallel FFT (Fast Fourier Transform) is actually executed on a clustered multiprocessor system with 32 processors to study real performance of three communication schemes, common memory, linear array and ring array. The speed-up of parallel FFT to a serial computer really depends on the number of data to be transformed and the number of processors. It is also proved that the performance really depends on the communication scheme of multiprocessor system.

A trial of statistical evaluation based on the probability distribution function is theoretically proposed in the actual case when a non-Gaussian type sound noise exhibits nonstationary properties due to a temporal change of statistical parameters. Next, when a nonstationary sound pressure wave of non-Gaussian type passes through a sound insulation system, the probability distribution of transmitted noise intensity is theoretically derived by reflecting the nonstationary properties of incident noise and the frequency characteristics (or impulse response) of the noise insulation system. The proposed evaluation method is employed for an aluminum single-wall whith nonstationary Gaussian type incident noise and compared with experimentally observed values. All results show good agreement with experimental results.

The waveform moments are used for analyzing the various types of signal waves. The ordinary moments can't properly express the characteristics of the given signal sequence, because they are calculated with simple summation formulas of the original signal values and the positive and negative values may cancel each other. So we define the power moments which utilize the square of signal in order to prevent the above defect. Moreover, the cross moments which evaluate the product of signals are derived as more generalized definitions. The cross moments can evaluate not only the characteristics of signal power but also the relations between two signals. Further, we extend the cross moments to the blockwise cross moments and the multicross moments, and then derive the corresponding expressions in frequency domain. Finally, we illustrate the usefulness of the cross moments with some application examples.

In this paper, we propose two parallel M/M/1 queueing systems with dynamic routing under a threshold-type scheduling. Arriving customer is sent to the buffer of the faster server as far as the difference of two queues does not exceed a threshold value, but it should be sent to the buffer of the slower server if and only if the difference of two queues exceeds a threshold value. It is shown that as the difference of the two service rates becomes large, the two parallel M/M/1 queueing systems with dynamic routing under the threshold-type scheduling can reduce the mean queueing time of costomers in the system compared to the conventional nonthreshold-type one. It is also found that as the asymmetry in the two service rates increases, the optimum threshold value which minimizes the mean queueing time should be large to reduce the number of queueing customer in the buffer of the slower server. On the other hand, as the traffic intensity becomes large, the optimum threshold value tends to be small in order to increase the total service rate.

An approximation method for analyzing open queueing networks with nonpreemptive priority is presented. This method takes into account the variability of the arrival processes and service times for each node in the network, and hence a general class of non-Markovian models can be analyzed. The approximation procedure has following three steps: (1) the separation of the two-class priority network into two single-class networks, (2) the analysis of the flow rates and variability parameters of internal arrival processes for each class network, (3) calculation of traffic characteristics for each node in the single-class networks. For the first step, a two-parameter virtual server method is established, in which the arrival processes and service times are approximately characterized by the first and second moments. The second step is based on the decomposition method for single-class non-Markovian networks. At the third step, traffic characteristics for the entire networks are obtained by assuming that the nodes are independent. It is possible to quickly analyze large network models by this approach, because the procedure does not require iteration or a great amount of calculation. Numerical results are given and validated by exact and simulation results for a tandem network model and by simulatons for the basic component models of complex networks. Also, we apply our approximation method to an end-to-end delay analysis for a packet-switching network for voice and data. The results indicates that the accuracy of the approximation method is sufficient for practical use.

A distributed power supply system with highly efficient characteristics has been developed to replace conventional centralized power systems in telecommunications offices. In this system as in conventional centralized power systems, a stand-by energy system composed of batteries and engine-generator sets is used as a back-up system for the commericial power lines. However, the distributed power system, except for the engine-generator set, is installed in the telecommunications room. Thus, the floor load limitation is an important design factor. This paper describes a stand-by energy system design that considers floor load conditions and is based on reliability analyses. These analyses show that the distributed power system should be designed as follows: (1) Using a fully duplicated engine-generator set. (2) With battery reserve time reduced to one sixth that of conventional centralized power systems. (3) With transfer switch unavailability designed to be less than one half of the existence switch unavailability.

This paper describes a new control algorithm for a transversal filter in a cross polarization interference canceller (XPIC) for digital microwave radio systems. The proposed new algorithm includes the shift bit select (SBS) method whose abjustment function of optimal correlation provides complete adaptability to the severe propagation condition changes. High speed control response under fading, as well as stability under steady state, can be achieved. Moreover, by combining the SBS method with the maximum level error method, wide acquisition range and short acquisition time can be realized. An analysis of convergence properties of the XPIC is presented, and using a 256 QAM signal, the availability with the SBS method experimentally confirms the superiority over conventional algorithm.

A new recognition method for partially occluded two dimensional objects is described in the present paper. Several vectors named PSVs, Partial Shape Vectors, are proposed as local features of an object and used for recognition. Two dimensional objects similar to a given model are extracted by this method. PSVs denote turning angles at an arbitrary point of the contour line of an object. In the recognition process, PSVs of each picture point in the given image are obtained and compared with PSVs of feature points which were set on the model. From coordinate values of the candidates for the feature points, objects similar to the model are extracted.

A framework for the visual representation of three-dimensional free-form curved surfaces based on a special class of surface curves, herein called the surface structure curves, is in progress. By analyzing their properties, we attempt to construct a basis for describing the topographical structures of curved surfaces which give a global description of the surface geometry. Surface structure curves are a set of surface curves defined by using viewpoint-invariant features-surface curvatures (and their gradients and asymptotes) from differential geometry. From these surface structure curves, surface sketches by means of the topographical structure of ridge lines, valley lines and the enclosing boundaries of bumps and dents can be inferred. This paper proposes a view-point invariant representation scheme which provides a smooth surface sketch",--a natural parameterization of free-form curved surfaces. We define three types of surface structure points and five types of surface structure curves in terms of zero-crossings, asymptotes and gradients of the Gaussian and mean curvatures. We discuss their properties and usefulness in edge based segmentation and description of free-form curved surfaces. Some examples of surface sketches using the surface structure curves are shown.