Application of Dedicated Fuzzy Hardware in Engineering Problems A.B. Patki
Aman Nijhawan
Deepali Aneja
Sankalp S. Parihar Abstract
Dedicated fuzzy hardware is required for a large number of engineering applications keeping in mind the myriad of problems at hand some of which have been discussed in the introduction. The need for dedicated fuzzy hardware is on rise keeping in mind the requirements for outsourcing boom which cannot be matched by routine software solutions. The paper discusses the fuzzy hardware implementation and a dedicated chipset. In this context, two components of the chipset viz., Fuzop and the hedge generator, have been studied in detail and a design consideration of fuzzy ALU has been examined. Further application areas of such dedicated fuzzy chipsets are also discussed. Index Terms-- Fuzop, Fuzzy ALU, Fuzzy Logic Chip Set, Hedge Generator. I. INTRODUCTION
T
HIS paper is based upon the work done in the area of Dedicated fuzzy Hardware and its implementation as a part of our internship project at Department of Information Technology, Ministry of Information Technology, Government of India. Fuzzy Logic is a problem-solving control system methodology that lends itself to implementation in a myriad variety of systems ranging from simple, small, embedded microcontrollers to large, networked, multi-channel PC or workstation-based data acquisition and control systems. It can be implemented in hardware, software, or a combination of both. This approach to solve problems mimics how human brain takes decisions based on imprecise and noisy information making use of the fact that some problems cannot be solved by using the crisp set approach. According to Daniel Pink the author of “A whole new mind”, right now we are experiencing a transition from Information age to Conceptual Age and in this scenario more emphasis is being paid towards the implementation and applications of soft
Ashish Aggarwal
computing technologies. This was our major motivation towards realizing this project. Fuzzy Logic can be implemented in hardware, software, or a combination of both. Till now software emulation was the most common approach to implement fuzzy logic based systems. Though it has advantages of Quick Design and Low cost but it has an inherent disadvantage of low speed so it cannot be used in real-time applications which require fast control. This problem was solved by general purpose fuzzy chips but using such general purpose hardware created another issue, the circuits realized using these chips could not meet all the requirements and external circuitry this made the system bulky and increased the power requirements considerably thus this kind of hardware was not suitable for real time embedded systems. To solve this problem of requirement specific fuzzy chips or dedicated fuzzy hardware is required. This type of hardware is suitable for applications which involve Hard Timing constraints and Low Power Consumption This paper describes the design of a fuzzy chipset which could be plugged into existing PC boards as an add-on card and later optimized into SoC architecture to provide special purpose fuzzy processing.
II. FUZZY HARDWARE DEVELOPMENT Fuzzy systems based on dedicated digital hardware can deliver much higher performance than those based on general purpose computing machines. The simplicity and versatility of some successful fuzzy inference algorithms, the advent of high density user programmable logic devices, together with powerful EDA tools, make dedicated digital fuzzy hardware a feasible solution for implementing high performance fuzzy systems. The fuzzy logic systems can either be implemented by digital hardware or analog design.
However, digital hardware has a marked advantage of system integration with existing PCs and digital systems as an add-on card. Various fuzzy modules can be generally implemented in 3 ways: (1). Software on standard computers, (2). Programmable fuzzy chips and (3). dedicated fuzzy hardware. One of the major drawbacks of solution (1) is that it runs relatively slower in comparison either (2) or (3). Since speed is a major consideration in most of the applications, they propel the development of dedicated fuzzy hardware. Dedicated fuzzy hardware becomes increasing important in domains where there are particular cases involving: 1. hard timing restrictions: optimization possibilities for systems with fixed hardware are very limited. 2. mobile applications: high power efficiency is required due to absence of programming overhead and limited memory considerations. 3. high production volume: decrease of chip area.
5. 6. 7. 8. 9.
Standard Hedge Generator Chip Standard Membership Function generator Chip Error Correction Chip Fuzzy Arithmetic Unit Intersection, Union and Complement Chip for Type-2 Chip
Dedicated fuzzy hardware development can be divided into 2 main categories. (i). Fuzzy Processor, & (ii). Dedicated Fuzzy Hardware (i.e. Fuzzy Logic Chip Set) A fuzzy processor is an application specific instruction processor on the lines of RISC processor, along with a neural processing chip as co-processor. It contains set of selected instruction to support certain fuzzy inference algorithm. It yields much higher performance than a standard processor and maintains a degree of flexibilty via the programmability feature. A fuzzy logic chipset as the name suggests, is a family of generic chips and depending upon the type of application to be selected, a particular array of chips can be assembled and configured to work together in an environment. The chipset which will have TTL compatible input output can be mounted on a printed circuit board and depending upon the type of operation to be performed, can be installed in a PC as well. A typical chipset has the following components (Fig-1): 1. Fuzzy Operations Chip 2. Fuzzy Inference Chip 3. Fuzzifier Chip 4. Defuzzifier Chip
Fig-1 III. DEDICATED FUZZY HARDWARE Out of the 8 chipset components 2 have been selected for further study and examination, viz. Fuzzy Operations Chip and Hedge Generator Chip. 1. Fuzop: A Fuzzy Operations Chip performs one of the many selected fundamental operations, i.e. Tnorm and Tconorm and Fuzzy Negation on two input values which can be any of the three functions from Drastic, Logical and Bounded. While a fuzzy negation performs f(a)= 1-a, where ‘a’ is the input value. Logical function: For Tnorm a^b=min(a,b) For Tconorm avb=max(a,b) Bounded function: For Tnorm a b=0 v (a+b-1) For Tconorm avb=1 ^ (a+b) Logical function: For Tnorm a^b =a, if b=1
=b, if a=1 =0 if a,b<1 For Tconorm avb=a, if b=0 =b, if a=0 =1, a,b<1 Where a,b are two inputs in the Fuzop chip. 2. Hedge Generator Chip: A basic hedge generator chip produces output corresponding to inputs in the chip. The basic hedges are concentration, dilation, extremely and indeed. Basic definition: Concentration is: CON(a)=a2 Dilation: DIL(a)=a1/2 Extremely: EXT(a)=a3 Indeed: IND(a)=2a2 if a<0.5 and =1-2*(1-a)2 if a>0.5
IV. APPLICATION OF DEDICATED FUZZY HARDWARE 1. Multiplexing of BPO Infrastructure. The applications of fuzzy hardware can very well also be extended to Business Process Outsourcing establishments. Business Process Outsourcing (BPO) is generally a cost saving and effective measure that helps an establishment to perform a large variety of secondary jobs that a company requires. The most common examples of BPO are call centers, human resources, accounting and payroll outsourcing. In the present scenario, BPO infrastructure is being used in absolute terms with respect to consumer community. Presently the jobs carried out in BPOs are in a more of an ad-hoc manner, with no such dynamic load sharing methods implemented to efficiently reduce the workload. Hence there is a need to think on the lines of relative BPO houses, where scheduling of jobs is carried out in a manner corresponding to the demands as well as the feedback of the user. The most plausible and logical yet novel idea seems to be the multiplexing of BPO infrastructure on the basis of geographical locations to provide and serve a large consumer base and user community. This architecture will help and enable the production of secure systems which can further enhance data security and enable safe data transfer. The OSI reference model for the TCP/IP network protocol can be extended to accommodate such multiplexed BPOs. Recourse allocation should be
carried out using platform independent Integrated Development Environment (IDE). Algorithmic steps for BPO Multiplexing. 1. Create a global Fuzzy Map (FMAP) of service centres available to the customers, mapping clients and customers to the multiplexed infrastructure based on membership function value. 2. Manage metadata through proxy servers that reference the locations where data blocks of different files are placed in an encrypted format. 3. Maintain a cache block so that some permitted, registered and trustworthy requests can be handles without accessing the secondary storage like machine disks and other resources directly. 4, Maintain cache consistency metadata, i.e. version information so that obsolete older version data is still preserved but not written onto; however allowing users to read from this data. Multiplexing of BPO infrastructure can lead to a far more secure, trustworthy and reliable BPO setup. With a large number of companies today off shoring their work to BPOs, it becomes imperative that an effective strategy and setup is chalked out to make these BPO stations far more reliable and efficient. V. FUZZY ALU The basic ALU of the processor is assisted by a Fuzzy ALU which helps in executing the instructions faster. The Fuzzy ALU plots a spread of the membership required function of the instructions in the processor. Further work is underway on Fuzzy ALU. VI. REFERENCES 1. 2.
3.
4.
5.
Donald H. Hung, “Dedicated Digital Fuzzy Hardware”, IEEE Micro, 1995. Silvio Triebel, J. Kelber and Gerd Scarbata. “Facta Universitatis”. Series: Electronics and Energetics Vol. 11, No. 3(1998), 1-8. H. Watanabe, “RISC Approach to Design Fuzzy Processor Architecture”, Proc. Int’l Conf. Fuzzy Systems, IEEE, Piscataway, N.J, 1981, pp 120-127. AB Patki, “Fuzzy Systems :Technology Mission Approach”, Technical ReportDE/NMC/935, May 1993. AB Patki and UP Phadke “Technology Watch: Fuzzy Logic based Hardware Systems”, Electronics Information and
Planning Journal, Vol. 21 No. 4, January, 1994, pp-214-217.