Abstract:This paper puts forward the architecture and organization of a high performance prolog-oriented intelligent machine system，which comprises a conventional computer as its host. A powerfull prolog processor supported by a special hardware is its back-end machine. Although the architecture of the system is simple，the performance is very high. Simulation results show that the performance of the system is about 500 KLIPS for logical inference and 17 MIPS for numerical calculations.

Abstract:Using the queueing theory, the queue model of memory with multibank in tightly-coupled multiprocessor system is constructed. The analyt1c expression of memory utility bandwidth is given by analysing the busy period of a bank. The relationship among the rate λ of the accessing memory of processors，the serving rate μ of a bank and the number m of banks is studied under the condition of keeping the memory utility bandwidth maximum.

Abstract:This paper gives a decision-tree algorithm acquiring many kinds of knowledge. The algorithm depends on the given priority and type of attributes to acquire the classified type of knowledge. In order to ensure the effectiveness of the acquired knowledge，a kind of evidence support degree to evaluate knowledge is proP9sed according to Cohen's inductive probability，and a kind of knowledge refinement method is given correspondingly.

Abstract:In artificial inte11igent，analogical reasoning is an important field of study，and similarity criterion is a significant concept of analogical reasoning. The paper gives a descriptive framework of analogical reasoning, and a definition of similarity criterion. Based on term rewriting technique，the paper proposes a verification method of the simi1arity criterion.

Abstract:At present the array-array (matrix-matrix) opreation has been introduced into parallel programming languages，such as Fortran 8x. This makes compiler implementation on serial machine difficut. This paper proposes the concept of “serialization” and gives the algorithm about serial opreating.

Abstract:The rational and Chebyshev approximation polynomials of tangent are given. The acceleratd computational methods of the Chebyshev coefficients are also presented. The experimental results on the YH-l super-computer show that using the coefficients with this accelerated，method. the accuracy can be increased by 30% compared with the original confficients.

Abstract:This paper introduces a method of the realization of the distributed programming by combining programming language C with distributed library function in the robot force-feedback compliance controlling multiprocessor system. The form of the distributed library function，the multiprocessor operating system function supporting the design of the distributed library function，and some problems which have to be emphasized in the design process are described. A simple distributed program example is also given.

Abstract:A new computer architecture-MIMD multimicroprocessor parallel processing system used in robot force-feedback compliance control-is described from the views of system hardware and software. The system is designed for fulfiling a practical research project. It can efficiently and parallelly process the multitask in the robot force/position mixed control. The realization of the system will make the servo-control period shorten to 5 ms and meet various needs of robot control tasks.

Abstract:Some techniques used to enhance the performance of an intelligent machine is described. The performance of the machine is simulated. Simulation results show that the average inference speed of an intelligent machine is about 500 KLIPS，which is twice as fast as Japan's PSI-Ⅱ and 2 times faster than PLM developed by Berkeleg College of California Univ. in U .S.A

Abstract:MT-1 is a database management system based on dBASE-Ⅲ. It has a graphic language for data description and manipulation. It is a useful tool for the development of MISS. It is superior to dBASE-Ⅲ in many aspects，such as graphic user interface，data dictionary，views，data integrity and security，etc. This paper describes the architecture of MT-l，and discusses some issues of its design and implementation.

Abstract:A language extension method through the transformation from source language to intermediate language is presented in this paper. This method has been used in the implementation of vector statements in YFT77 and it has been proved to be efficient and practical.

Abstract:This paper describes a double compression method with characteristic block compression，and then Huffman block encoding for computer graphics in detail. The method has been implemented in the YHHT graphic software. It has been shown by applications that the method is highly efficient.

Abstract:The paper discusses the existence of optimum(n,2,w) binary constant weight error-detecting codes. A succinct proof is given for the conclusion made by prof. Wang xinmei (china science,vol.11,1987) when n is even. The author also gives a proof for the conjecture of the wang xinmei’s paper.

Abstract:This paper presents the concept of recurrence sampling of m-sequences. The complete recurrence sampling of the m-sequence with a prime number period is studied and solved. In the case of composite number period，some useful conclusions are achieved.

Abstract:This paper discusses the periodicity of feedforward sequences.A general method to get the longest period of feedforword sequences is put forward.

Abstract:In order to improve the stability of a filter under weak observable condition，the nonlinear optimization method is used for recursive fi1tering and a damped least squares iterated fi1tering algorithm has thus been derived. The form of information square root filtering of the method is presented.

Abstract:The problem of adaptive detection of transient signals is studied. Based on the idea of using vector length feature of autoregressive parameters，a method of estimation of parameters is presented. This method is of good performance of adaptive noise control and it can make up the deficiency of LS and SVDT methods. Theoretical analysis and simulation results show that this method is of good performance of autoregressive parameter estimation and transient signa1 detection.

Abstract:How to maintain the false-alarm-rate constant of radar detecting system is one of the important problems in modern radar technology and signal processing technology. In order to obtain constant-false-alarm-rate in the complex clutter surrounding，radar detecting system must have a constant-false-alarm-rate processor (CFAR Processor). The AC-CFAR processor discussed in this paper is a new processor. Its CFAR threshold is implemented by selecting a mean value which approaches the detected distance-cell value. On the basis of simple hardware-implement，the AC-CFAR processor has an adaptive performance which can timely reflect the changes of external clutter，and a very good perfermance of anti-multitarget effect and antic1utter edge effect as well. It can also resist the effect ofb“ig target eat small target” and reduce CFAR loss.

Abstract:The paper discusses the technology of speed measurement of projectiles with FFT method. This method is superior to that of the traditional analog filter speed measurement. It can measure the speeds of many projectiles in the antenna. The signal，which was filtered by the digital filter，has higher signal-to-clutter ratio and noise ratio. Thus，the distance of measuring speed can be increased greatly. The simulation results show that the higher accuracy can be obtained by using FFT technology.

Abstract:A microcomputer inferface with multiple functions is presented. It 1s a kind of test and control means with strong function and can be applicated in the fields of microcomputer control，signal test，circuit self-test，fault diagnosis and digital signal generation with much practica1 value.