Uzman sistemler ve ulaştırma alanında kullanımları

Abstract

Yapay zekanın bir uzantısı olan uzman sistemlerin popüleritesi her geçen gün artmaktadır. Uzman sistemler, yapay zekanın finansal açıdan lokomotifi durumundadırlar. Çok az bir geçmişe sahip olmalarına karşın elde ettikleri başarı hiç de azımsanıcak gibi değildir. Bu başarıyı elde etmelerindeki en büyük pay sahibi, kuşkusuz uygulamaya yönelik olmalarıdır. Uzman sistemlerin temel amacı, insan tecrübelerine dayanan, muhakeme ve sonuç çıkarım aşamalarında bir takım prosedürlerle birlikte insan karar verme yeteneği gerektiren bazı aktiviteleri bilgisayar yardımı ile gerçekleştirmektir. Teşhis, planlama, tanıma, yorumlama, tasarım gibi konularda insan davranışlarını taklit eden uzman sistem çalışmaları yapılmaktadır. Geçmişi son derece az olan uzman sistemlerin ulaştırma alnında kullanımı son 15 seneye dayanır. Fakat bu konuda da oldukça değerli ve yararlı çalışmalar yapılmıştır. Ulaştırma sistemlerinin analizinde, alt yapı kontrolünde, bakım-onarım işlemlerinde, arıza teşhisinde, ağ kontrolü operasyonlarında gibi bir çok konuda yapılmış ve yapılmakta olan uygulamalar vardır. Yapay zekanın bir kolu olan bulanık mantık, belirsiz durumlardaki çıkarım işlemleri için büyük olanaklar sağlamaktadır. Ulaştırma problemleri de günlük hayatın içinden kesitler içerdiği için bir çok belirsiz durum içermektedir. Son yıllarda bu belirsizlikler, 1960'ların ortasında ortaya atılan Bulanık Mantık Teorisi Yardımı ile çözülebilmektedir. Konu ile ilgili çeşitli araştırmalar yürütülmektedir ve bir çok ulaştırma probleminin çözümünde önemli mesafeler alınmıştır. Bu çalışmada, yapay zeka ve uzman sistemler hakkında genel bilgiler verildikten sonra, ulaştırma alanında kullanımları ele alınmıştır. Daha sonra ise belirsizlik ve Bulanık Mantık Teorisi'nin ulaştırma problemlerindeki yeri anlatılmıştır. Son olarak ise, örnek bir uzman sistem tasarlanmış ve değerlendirme yapılmıştır.

The founding of artificial intelligence as we know it today is often traced to a small summer meeting of researchers at Darthmouth College in New Hampshire, USA in 1956. In the decades following "Darthmouth Conferance," artificial intelligence researchers have developed computational models and computer programs that attempt to solve tasks in a way that would be considered intelligent if performed by human. A conceptual breakthrough in artificial intelligence research was occured with the recognition that extensive, high-quality knowledge gives a computer program is problem solving power in ill-structured or nonalgorithmic, narrow, problem domains. The subsequent development of knowledge-based expert systems founded on this concept has consituted one of major successes of artificial intelligence. Expert systems have recently become popular and attracting more and more attention. The high quality performance achieved by some experts in areas perivously not considered practical for computational solutions has lead to great interest for many different disciplines. Most expert systems use a subset of techniques from general area of computer science research known as artificial intelligence. The techniques used in expert systems may be what is needed to biridge the gap between classical operational research modelling and human decision making processes. A knowledge-based system is a computer system thet attempts to replicate intelligent activities of specific human experts. Human experts make decisions and recomendations, such as adjusting temperature controls in manifacturing plant. They also assist and train others to do tasks and to make decisions. A knowledge-based system enables users with a problem to consult the computer system as tehey would an expert human advisor. Like a human expert, such a computer system extract needed information from the user by asking questions related to problem during consultation. It can also answer questions asked by user about my certain formation is needed. It can make recomandations regarding the problem or decision at the end of the consultation, and it can explain the reasoning steps it used to reach its calculations. In some situations, it can analyze a problem and take corrective action directly. Knowledge-based systems are sometimes classified by the area to which they can be applied, such a medicine and chemistry. In other instances classified by generic problem areas they concern with, such as those given below: Diagnosis: Infers system mulfunctions from observatins. Types of systems are medical, electronic, financial analysis, auditing, machine repair, and etc. Monitoring: Compars observations in order to identify variations. Types if systems are management control, nucluer power plan, and etc. Debugging: Prescribes remedies for mulfunctions. Type of system is compuer software. Repair: Executes a plan to administer a prescribed remedy. Types of systems are automobile, computer, telephone, and etc. Instruction: Diagnoses, debugs, and corrects student behavior. Types of systems are tutorial and remedial. Control: Interprets, predicts, repairs, and monitors system behaviors. Types of systems are air trafic control, battle management, manifacturing process control, and etc. Prediction: Infers likely consequences of given situations. Types of systems are weather forecasting, crop estimation, financing forecasting, and etc. Interpretation: Infers situation descriptions from sensor data. Types of systems are speech understanding, image analysis, surveillance, mapping, and etc. Design: Configures objects within situation constraints. Types of systems are circuit layout, budgeting, automatic program generation, and etc. Planning: Developes guidelines for acting. Tyoes of systems are strategic planning, process scheduling, military planning, and etc. Classifastion: Prescribes categories for given sets of criteria. Types of systems are planning, scheduling, layout, remedial, auditing, forecasting, and etc. Expert systems are suitable for aplication to specific problems that are more amenable to treatment on the basis of rules and reationships rather than by numerical calculations. Further, these problems may be considered on the basis on incomplete or even conflicting information. Superficially at least, it would appear that transportation would be one area of study where knowledge-based system woul be applicable. In particular, there should be many possibilities for using knowledge-based system in processes of planning and operating transportation systems to enhance the quality of ecision making. At the same time, many problems of transportation planning and management can involve the use of complex mathematical models (e.g. for transport network analysis, travel demand estimation, network capacity, and travel cost estimation). Knowledge-based system strategies may thus needed to be considered in conjunction with algorithmic models of travel behaviour and performance. There are number of reviews of the possible applications of knowledge based systems in tarnsprotation planning and engineering, for example. Takallou (1985), Bonsall and Kirby (1986), Ritchie and Harris (1987), and etc. A notable feature of these reviews is the relatively few operational knowledge based system reported, especially in the areas of planning and operations. Ritchie and Harris (1 987) provided a useful cross- classifaction of operational and demonstration expert systems by subdividing the field of transportation engineering into following categories: planning, design, operation and control, management, maintenance, and rehabilitation. Their classification table indicates a predominance knowledge based system in the areas of design, management, and maintenance. In some respects this outcome is matter of necessity. As indicated by Ritchie (1987), there are significant forces that have required an attention to existing and impending problems in (say) pavement management, where large-scale rehabilitation and reconstruction of existing road networks is needed, and where a significant loss of human expertise is occuring through old age and retirement. Expert systems research in the transportation field is comparitavely recent. However, significiant research contribitions have been made, and considerable processes has been achieved, in this first ten or so years. These are the sign of what the future will bring. Increasingly, we can expect to see the maturing and transition of expert system technology. Moreover, providing adequete research support is comitted, the comming years promise to be especially exciting and challenging ones for expert system techniques to be applied in the integrated solution of difficult transportation problems. Artificial intelligence would have made much more progress toward its goals if it had not committed itself exclusively to symbol manipulation and first-order logic. This commitment has made artificial intelligence somewhat inhospitable to methods that involve numerical computations, inculiding neural and fuzzy methods, and has severely limited its ability to deal with problems where we can not benignly neglect uncertainly and imprecision. Most real-world problems fall into this category. The theory that can express and solve the problems that are uncertain was invented by a scientist whose name is Lotfi Zadeh in 1 965. That theory brings a new aspect to many problem in uncertain domain. Because transportation problems come from the real world, many side of these problems are not so clear. So, this theory coul be used to solve tasks in this manner. Many research are being made on this new and challenging subject. In this thesis, firstly, basic concepts of artificial intelligence was expressed. The history, today and future directions of artificial intelligence was explained. After these, advantages and fields of artificial intelligence was indicated. Then, some applications are made by using artificial intelligence techniques were described. At the second step, expert systems that are the branch of artificial intelligence was explained. Advantages and disadventages of expert system and the areas they are used discussed in this section too. By the help of this discussion, the decision about when, where, and how to use them could be given much clearly. The section after, the role of expert systems in transportation was discussed. Questions like, why and where do expert systems are used in transportation were answered. Some expert systems are developed to solve particular and significiant transportation problems were explained and detailed too. Then, Fuzzy Logic and Fuzzy Set concepts are considered with uncertainity. Many exmaples and applications were give to illustrate what is understood from the concepts of fuzzy. After these, an example that is related with airtraffic control with fuzzy logic were given. This example is the best one that can express the importance and illustrate the role of this concept. At the end of the thesis, an aplication that is written with VP Expert shell program was developed. This program is not so complex and not have specific task solving power. Just the aim of developing of this sytem is to indicate the obtaining a rule base structure an dto show the problem solving approach in knowledge-based systems. In this manner, program could be succesfull to illustrate an expert system's rule-base structure. Inference method of this system is backward chaining. According to this method, firstly a goal is selected ans sub-goals are used to reach the major aim. Consequently, expert systems are new, efficient and challenging technology. In the future we will be able to see quite importatnt applications developed by expert systems. So, we can use them in solving of many transportation problem. But the main important thing is to decide when, where, and how to use them. To overcome this problem, we have got to practice too much. According to all this expressions, we should select our objective function true. Otherwise, we could not reach our aims even we try so much.