Wednesday, June 08, 2005

A Global Web-Based Approach to Artificial Intelligence

The World Wide Brain :


A Global Web-Based Approach to Artificial Intelligence


Introduction

This paper will discuss an approach to Artificial Intelligence that is based on the Internet and the World Wide Web – the World Wide Brain. In short the World Wide Brain is a merger between the fields of AI and Networking to evolve the WWW into a global web-mind. I will begin by discussing some fundamental topics; I will discuss why the World Wide Brain is the natural and likely progression of the WWW and Artificial Intelligence. I will talk about why this road towards Artificial Intelligence is not science-fiction, how it is marketable, and discuss its practical applications. I will follow this up with how the World Wide Brain could be implemented in the foreseeable future and the possible methods used to do so. I will conclude this discussion with some problems and philosophical issues that would come from the evolution of the World Wide Brain.

Much of this paper discusses the visions of Ben Goertzel and his opinions of the World Wide Brain. Some of my own opinions differ from his however, so this paper is both research and introspection of my own outlook. Other resources of information will be presented in the references section.

Why the World Wide Brain

Artificial Intelligence has thus far eluded modern science. The baby steps we have taken in the field are very small compared to the exponential leaps that computer hardware has taken place in the past few decades. There are many reasons for this but the fundamental problem to Artificial Intelligence is that it is a very difficult topic to handle since it has a great deal of complexity. AI is primarily linked with Computer Science, but it has many important links with other fields such as Math, Psychology, Cognition, Biology and Philosophy. Computers have a great deal of trouble understanding specific situations and adapting to new situations. They have little in the way of abstract thought, high-level deliberative reasoning and pattern recognition. These attributes are key for Artificial Intelligence and as such AI development looks to be a project of epic proportions.

While the above may be true, it is not to say that it is impossible. Spaceflight has shown us that anything can be done given time, labor and a limitless supply of disposable income. Unfortunately Artificial Intelligence has been given little of any of these things – so what is one to expect? Artificial Intelligence will likely go no-where until public enthusiasm is generated for the subject. Until then, it will be the field of Japanese engineers and American software developers with an incidental interest and similar correlation of problems to solve. Indeed, it is a poor method if you wish to develop something with results.

Public support is being generated for Internet in ever-rising numbers however. Businesses are forced to adapt to E-Commerce or die. This monster of the ether is indeed thought of as an entity of its own, and for good reason. It has defined the last decade and it will likely continue to evolve for quite some time. We have already seen the major changes from primitive BBS boards, to HTML page and then from scripted CGI to an Active Web of XML and Java. A good deal of money is invested in the internet to prod along the evolution of the Web and this can likely be used to fuel a merger between Artificial Intelligence and the Internet. An active, more efficient, and a more useful Internet is very desirable and AI can definitely be applied to this type of functionality.

The Internet has several things going for it in terms of its usefulness as the launching pad of an Artificial Intelligence. First and foremost, the first company to integrate an AI will have a sizable advantage (and public interest) over his competitors. This means money going into the field wither they know it or not. Secondly the Internet is an acceptable platform for AI in the current generation. People may not accept a free-willed robot roaming around their apartment but they will likely dump their life savings into the WWW if AI enhanced super-porn becomes a reality. Not to mention what AI can do for stock trading, information filtering and organization, games and a host of other useful functions. Acceptability by the masses is extremely important in an age where many technological advances have titanic promise, but the elder-folk (those in power) scorn their use – cloning, embryonic research, the human genome utilization are just a few to mention.

The third advantage of the WWW is that it already closely resembles the human brain, in many instances, much more so that any other piece of technology created. Packets of information fire like neurons from one synapse to another. Sprawling networks and clusters use a collective effort to problem solve. Huge stores of information are ready to be recalled at a moments notice. While it is true today’s computing machines are hardly comparable to the human brain on their own, they are comparable when used in large groups.

The internet as a whole handles more information several orders of magnitude higher than anything before it, including our brain. SETI@Home (http://www.setiathome.ssl.berkeley.edu/) is an excellent example of a collaboration effort of many thousands of personal computers clustering their processing power towards a single task. This project has been going on for nearly a decade and it is has had some amazing results analyzing huge amounts of data. Even with several clusters of computers at their disposal, the SETI program could not hope to match the nearly unlimited processing power of the general public. Considering the program is operated nearly entirely by good-will it has done an amazing job and will set a good precedent for similar mass-user projects including an AI global web-mind project.

It is all but inevitable that new kinds of computers will not limited by physical distance and will only increase in speed and efficiency exponentially. As a group, the Internet will likely have (or already has) the processing power needed to operate an Artificial Intelligence. The currently limitations are on bandwidth and more importantly, organization and software. These limitations are fairly minor compared to the insurmountable problems one would have by creating an AI on a single machine. Where does one get the resources necessary to even begin to develop such a system? Why would it be useful? Surely it is cute to make such a thing, as IBM has made a score of chess playing machines as corporate promotions. However, in reality a cluster of machines donating a little bit of processing power can do the same thing as a far better price.

That said, the Internet as the platform looks to be a promising and perhaps natural progression to develop AI. The steps to developing AI would be useful unto themselves in our age of interactive web experiences. Organization and a cohesive development effort would be all that would be needed to begin a profitable and useful evolution from the WWW to the World Wide Brain.

Evolution and Implementation of the World Wide Brain

Before I discuss how we would evolve the WWW into the World Wide Brain, we should talk about what exactly we are evolving into. Ben Goertzel describes the World Wide Brain as follows:

“The global Web mind is what happens when the diverse population of programs and agents making up the Active Web locks into an overall, self-organizing pattern -- into a structured strange attractor, displaying emergent memory and thought-oriented structures and dynamics not directly programmed into any of its parts.”

Personally my favorite Artificial Intelligence definition is as follows, Artificial Intelligence is a branch of Science which deals with helping machines find solutions to complex problems in a more human-like fashion” (courtesy of AI depot http://ai-depot.com/Intro.html). This is remarkably similar to the above definition of the World Wide Brain/Global Web-Mind. Everything seems to boil down to making computers think more like us, or at least having a portion of them do so in a way that it is indistinguishable for us people.

Now that we have a definition of what AI is, I’ll simplify the process of going from the current WWW to the World Wide Brain into three main segments. The first achieving the necessary hardware backbone to make AI feasible and practical. Secondly I will talk about the first evolutionary step, the Learning Web. Then I will discuss the next evolutionary steps, the Thinking Web and Consciousness.

Hardware is generally the least of the problems with considering a World Wide Brain. As I have mentioned before, it is all but inevitable that new kinds of computers will not limited by physical distance and will only increase in speed and efficiency exponentially as time goes on. Hardware can certainly go nowhere but up in terms of speed and memory capability. Numerous development projects have great promise for breakthroughs in hardware attributes. The reason for this is simple, money. Computer hardware is a cash cow and we likely have yet to see the golden age of hardware come to fruitation. Because of this, one would surmise that the necessary hardware for AI would be a matter of time, if it is not already applicable considering the resources of the WWW. The World Wide Brain involves a global structure for its operation and is less limited by individual machine(s). I would think the bottle neck would be in terms of bandwidth and user willingness to devote a portion of their CPU time to an AI rather than the limitations of processing power or memory.

The most difficult aspect of the Hardware side of the World Wide Brain is the organization. One example I used before is SETI@Home, a single non-profit organization which allows the good-will of folks to assist in their efforts. A World Wide Brain may have none of these attributes unless it is properly organized. Should development start on a World Wide Brain project, first as separate Learning Web project and then into a more cohesive Thinking Web project - who is to say it is the only one. Competing AI projects who wish to use the same resources (bandwidth notably) could cause some interesting problems. I hate to think of possible AI wars competing against one another, but I suppose that several AI entities that war against each other would be the most human of all.

The route likely taken would be individual software projects that further emphasize the developing active web into a Learning Web. Down the road when AI development reaches a marketable position, a more cohesive collaboration would be devised called the Thinking Web, which I will talk about below.

Another useful hardware adaptation would be better incorporation of mutual integration between the PC and People. Natural language is one barrier, but I would say a computer that can understand what we are thinking would be more useful. Already been experiments in which people managed to steer images on a computer screen simply by thinking, moving an image across the screen. A device (neural interface or some such) that could literally read our thoughts would perhaps be better than any natural language translator; which are remarkably difficult to produce. After some of these cornerstones are in place larger and more involved AI projects could be worked on. The software evolution would look something as follows:

Pre-Web (-1990 Bulletin Boards, Email, User/News Groups)

Primitive Web (1991-1997) (WWW, Multimedia, Secure Web/Email, CGI)

Active Web (1998-20xx) (Java, XML, DB Integration, Knowledge Management)

Learning Web (Learning algorithms, Judgment, Associative learning, Self-Organization)

Thinking Web (Software Agents, Solving Complex Problems, Knowledge Discovery, Mutual Interaction, Understanding Speech/Thought)

Consciousness (Self-aware, Free-will, Self-changing code)

We are currently in the age of the Active Web, which is a web that uses real-time programs and exchanges a variety of information (pictures, text, sound) for a particular function. Java and XML are good examples of the active web. We are still in the young stages of the active web and tools will likely become more organized and interact better with each other and the user in the near future.

The Learning Web is the next evolution of the WWW. It will have some definite fundamental changes that lean towards AI. The web itself will organize the millions of pages of data to better suit the needs of the users. It will develop an associative learning capability to allow frequently data to be used together and become more strongly connected. In this way the Learning Web will develop methods to make judgments and predict outcomes to certain activities. This will also require some interaction on the users behalf so the Learning Web can keep track of the patterns and activities of each users. Perhaps something like super-cookies, piggy-back packets, or some other form of self-communication and interaction. The final stage of the Learning Web would be learning algorithms. This is mainly to immediately record patterns of usage and assimilates the collective wisdom of all people consulting the Web. This would be far more efficient than the crude indexing that we have now on Google or Yahoo.

The Thinking Web builds on these concepts and folds them into a cohesive framework to solve complex problems. A self-organized web could be much more easily searched and software agents could be run to do specific tasks or collect a criterion of information. A software agent would use your past patterns and its knowledge of your personality to perform the task/search.

For example, I could ask a software agent “what’s new today?” and it would immediately collect topics of news, weather, sports, stocks that are in my direct area of interest. Another example would be to ask an agent “patch my software”, and again it would update the drivers and general code for your PC. Nothing partially spectacular, but it still requires a great deal of technology to pull off, notably a thinking process to understand language, meaning, querying its amassed data and developing a process based n these results. Also it would be the precursor to ‘computer, write a program that does this’.

Another step of the Thinking Web is the ability to solve complex problems and to have its own knowledge discovery. Complex problems are slightly different than the above software agents because they would involve indirect information. If I ask the computer “I’m stressed out, what should I do?”, it would require something beyond merely searching for all possible information. The AI would have to make its own judgment calls and assimilate knowledge from its environment. Could the stress be allergies, weather, bad day at the job, and so forth. This brings forth another important phase of the Thinking Web, mutual interaction. In the example above the AI would ask you some questions so it could develop a better answer. Indeed the Thinking Web would be constantly collecting data all around the globe to better fulfill the tasks that are required of it. The AI would gain the capacity to automatically create new concepts, rules and models, and thus change its own way of thinking. With this, we are coming into the beginnings of a true AI and less into ignoble code that merely solves problems.

With the Thinking Web in place we are still missing a few important items of what some consider true AI however - Consciousness and Free Will notably. These would be the next step and it is questionable how would one go about handling this type of project. Thought-oriented structures and learning algorithms are a step in the right direction, but when would we classify something as ‘thinking’, having ‘consciousness’ or ‘free will’? These are very subjective terms to us and these topics are more ambiguous and are harder to achieve because of this. Many seem to think that if a Thinking Web were active, it could theoretically develop its own consciousness given time.

Conclusion

With consciousness many problems arises. Do we really want our tool to have consciousness and free-will? These are more philosophical questions and they are distant enough to be alien in our current frame of mind. We simply can’t not perceive most of repercussions of a free-willed global-web mind, and we can not make a very good argument in either direction.

From my research I have found the WWW and Internet in general to be a good foundation for an emerging intelligence. The web is already a rather large super-structure of information and it is indeed slowly evolving every day. It seems natural to me that the WWW will eventually develop at least some of the attributes I have talked about, and perhaps a true intelligence of its own. The steps to seem logical and productive on their own, which is rather contrary to many contemporary AI projects that try to tackle the problem as a whole with nothing marketable in the process.

References

· Vrije Universiteit Brussel, Brussels, Belgium (2001)_”From Intelligent Networks to the Global Brain Evolutionary” Social Organization through Knowledge Technology - The First Global Brain Workshop (GBrain 0)

Fuzzy Logic and its Practical use in Mass Transit Systems

Fuzzy Logic and its Practical use in Mass Transit Systems


 


Problem Statement

In recent years it fuzzy logic has become more and more prolific. Chances are that if you have bought any appliances from a washer and dryer to a coffee machine that it probably uses fuzzy logic in its day to day functions. Home appliances indicate the prolific use of fuzzy logic but I am surprised to read that fuzzy logic stable enough and trusted enough to operate highly critical machinery in a complex and ever changing environment – in particular recent mass transit systems such as the Maglev trains in Japan and Germany. The city of Sendai in Japan has its fuzzy control of subway since 1988, which is a good case study for this type of project. My paper will concentrate on why fuzzy logic is used in these types of systems (namely in Sendai), what problems needed to be solves in implementing fuzzy logic, and if the 16 years of fuzzy logic implementation in Sendia is seen as a breakthrough for Fuzzy Logic or perhaps illustrating flaws in fuzzy logic.

Definitions

To make this report as clear as possible, let me define two terms that I will refer to throughout this paper:

Fuzzy Logic: Fuzzy logic is a superset of Boolean logic dealing with the concept of partial truth. Whereas classical logic holds that everything can be expressed in binary terms (0 or 1, black or white, yes or no), fuzzy logic replaces Boolean truth values with degrees of truth which are very similar to probabilities, except they need not add up to 100%. This allows for values between 0 and 1, shades of gray, and maybe; it allows partial membership in a set. It is related to fuzzy sets and possibility theory. It was introduced in 1965 by Dr. Lotfi Zadeh of Berkeley.

Magnetic Levitation Train (Maglev) - A magnetic levitation train or maglev is a train-like vehicle that is suspended in the air above the track, and propelled forward using the repulsive and attractive forces of magnetism. Because of the lack of physical contact between the track and the vehicle, the only friction is that between the carriages and the air. Consequently maglev trains can travel at very high speeds with reasonable energy consumption and noise levels (systems have been proposed that operate at up to 650 kilometers/hour (404 miles/hour), which is far faster than is practical with conventional rail transport). Whilst the very high maximum speeds make maglev trains potential competitors to airliners on many routes, the costs of constructing the tracks are still high.

Definitions attainted at Wikipedia: http://en.wikipedia.org/

Maglev trains are not used in Sendia but they will be talked about in conjunction with traditional rail systems. Much of the early work was identified and set forth by Lotfi A. Zadeh, Ph.D., University of California, Berkeley. He described fuzzy logic as follows,

As the complexity of a system increases, it becomes more difficult and eventually impossible to make a precise statement about its behavior, eventually arriving at a point of complexity where the fuzzy logic method born in humans is the only way to get at the problem.”

This is an arguable point since very complex systems have been designed without the use of fuzzy logic. In particular I wanted to find material that talking about the pros and cons of fuzzy logic. For the most part this has been possible, but specific instances involving the Sendai Subway have been elusive. For the most part I will be relying on facts distributed about the Sendai project.

Origins of the Fuzzy Controller

Fuzzy Logic was first implemented in England in 1973 at the University of London. A professor and student were trying to stabilize the speed of a small steam engine the student had built. They had good equipment for the time but they had problems with the engine control. Engine speed would either overshoot the target speed or arrive at the target speed after a series of oscillations, or the speed control would be too sluggish, taking too long for the speed to arrive at the desired setting.

This experiment describes a common engineering problem and this is the heart of the fuzzy logic application – the machine was essentially not doing its job well because its understanding of its environment was poor. E. Mamdani, the project lead, had read of a control method proposed by Dr. Lotfi Zadeh, head of the electrical engineering department at the University of California at Berkeley, in the United States. Dr. Zadeh is the originator of the unorthodox designation ‘fuzzy logic’. Fuzzy logic was used Professor Mamdani’s engine controller and thus the first fuzzy controller was born. The controller worked very well and the speed of the engine reached it target in a third of the time with no oscillations in acceleration – simply a smooth movement from start to finish.

This fuzzy control would go on to serve in many practical applications include the ones used in the Sendai subway. In 1988 Hitachi turned over control of a subway in Sendai, Japan, to a fuzzy system. This is perhaps the most visible application of fuzzy logic, the Sendai Subway System of Japan, and it includes such devices as smart transmissions, braking systems, and camcorder focusing systems. Fuzzy logic is also used in planning traffic and predicting customer usage of subway facilities. Since then, Japanese companies have used fuzzy logic to direct hundreds of household appliances and electronics products. The Ministry of International Trade and Industry estimates that in 2002 Japan produced about $3 billion worth of fuzzy products. U.S. and European companies still lag far behind. Although the main focus of the paper is on the success of Sendai, I will talk a bit about the over all application of fuzzy logic as well.

Fuzzy Logistics

Fuzzy logic is controversial. It is widely accepted within the engineering and computer science communities but generally rejected by mathematicians and statisticians. Critics argue that it is unscientific by the standards of Karl Popper, since set membership values are not empirically verifiable. So is Fuzzy Logic fit for every day use when human lives are affected?

A good example is mass transit systems such as the Maglev and subway trains in Japan and Germany that widely use Fuzzy Logic. The city of Sendai in Japan has had fuzzy controls in it’s subway since 1988, and this will be an ideal case study for the practical use of Fuzzy Logic since it has been implemented for 16 years. My paper will concentrate on why fuzzy logic is used in Sendai, what problems needed to be solves in implementing fuzzy logic, and if the 16 years of fuzzy logic implementation in Sendai is seen as a breakthrough or perhaps illustrating flaws in fuzzy logic.

It is important to understand the strong and weak point of fuzzy logic to give an accurate assessment of the Sendai project. I think the most important aspect about fuzzy logic is that it is very tolerant on Imprecise Data. It can take vague information and give a reasonable suggestion for a response, much in the same way that people do. It is also conceptually easy to understand because it is based on linguistic terms. For example, if the weather is warm, wear cool clothing. Warm and cool a terms that we are familiar with, but is vague for a machine to understand. Fuzzy logic also is good with expert and common sense knowledge. It is good as universal approximation and fuzzy logic can model abstract non-linear functions.

Conventional approaches may be formally verified to work, where as fuzzy logic can not be empirically verified. This makes some scientist weary of fuzzy logic. Other methods are more direct and crisp precise models may be more efficient for certain task because they are very direct in how they move data around and make choices. Lastly fuzzy logic is not a cure all. It is not good for many tasks and it is an arguable point that fuzzy logic is not worth using at all. However it does seem to be a complementary choice for machine speed control, such as for the Sendai subway.

Fuzzy logic is used in many different fields. Fuzzy mathematics is used in many ways. Some techniques from fuzzy mathematics include fuzzy relation equations, group decision-making, abstract algebra, clustering methods, belief functions, fuzzy measures, evidence theory, sugero-integrals, abduction, automata theory, genetic algebras, and hyper graphs. The medical field is known for using many of these methods.

Fuzzy Expert systems can potentially be used in any expert system that would benefit from a non-Boolean set of functions and rules.  These include Medical Diagnosis, Stock Market Analysis, Mineral Prospecting, Weather Forecasting, and Politics. Fuzzy logic applications are used in many every day devices from toasters to coffee machines.  Interesting examples I have found include the battle AI for Lord of the Rings, , Noise Detection on Compact Disks, Air Control in Soft Drink Production, and of course Mass Transit Systems.
 
We should ask ourselves what the primary goal is of a fuzzy logic in the Sendai system.  The goal is fairly simple – to create a more efficient and cost effective and safe subway.  It is said that you can do far more with a simple fuzzy logic BASIC or C++ program in a personal computer running in conjunction with a low cost input/output controller than with a whole array of expensive, conventional, programmable logic controllers.  While this may be true one necessarily better than the other.
 
I found it interesting that some claim fuzzy logic can effectively go faster, smoother, and stop at exactly the right place in less time for a mass transit system such as a Maglev train and the Sendai subway.  Claims were made that trains could be stopped in exactly the right place, which was not possible with conventional controllers. I am skeptical since older trains in the USA and England can be seen to routinely stop very smoothly and crisply, with less than an inch of position error. This is because the track-side door had to align with the door of the train, within less than an inch of tolerance. Even the PRT at WVU has accurate stopping mechanisms (I have found no material that states the PRT runs with Fuzzy Logic). These transit systems could stop perfectly and consistently and these systems are decades old, some say far outdated. Although they do seem to stop at the right place, but only by slowing to a crawl and then creeping up the last few inches.   Once again we come across oscillations in acceleration and deceleration and this is wasted seconds. The PRT seems to slow within three meters of its target and stops within about two feet of its target before inching along.
 

Non-fuzzy engine speed should either overshoot the target speed or arrive at the target speed after a series of oscillations, or the speed control would be too sluggish, taking too long for the speed to arrive at the desired setting. When Fuzzy Logic is applied, you can achieve a sense of no-motion from start to stop. A fuzzy variable becomes a linguistic variable when we modify it with descriptive words, such as somewhat fast, very high, real slow. Speed is a fuzzy variable. Accelerator setting is a fuzzy variable. Examples of linguistic variables are: somewhat fast speed, very high speed, real slow speed, excessively high accelerator setting, accelerator setting about right, and so forth.

For the older trains, fuzzy logic had not been invented when they started running so I doubt Fuzzy logic has upgraded most of these systems. Part of the problem is that I have not actually have been able to ride a fuzzy logic train (to my knowledge) so I have nothing to compare the conventional experience with. A ‘smooth ride’ for example, it is a very ambiguous term, and a smooth ride is not necessarily our target goal. Efficiency and less energy consumption is our target goal. I have read a good deal of material that is contradictory on the subject. Some say fuzzy logic is clearly more efficient, some say that normal methods function identically, and surprisingly I have read arguments that traditional controllers are more refined to their task an perform better.

Fuzzy Logic in the Sendai Subway

One of the most successful fuzzy logic implementations is the control of subway in Sendai, Japan. The fuzzy system controls acceleration, deceleration, and breaking of the train. Since its introduction, it is said that it not only reduces energy consumption by 10%, but the passengers hardly notice now when the train is changing its velocity. In the past neither conventional, nor human control could have achieved such performance. The ten percent claim may be in dispute but the acceleration gradient is not. There is no doubt that the change in velocity is much better than a standard controller can produce, but the efficient claim is disputable.

The Sendai Subway is slightly over 9 miles and uses an overhead wire with 1067 mm gauge and 1500V DC. The trains operate every four to six minutes depending on the cycle of the trains, the time of day, and the dictations of the fuzzy logic traffic. The line has 17 stations, and most are underground. The line contains some curves but the majority of it is in a straight line. The price for the line varies from 200 Yen to 350 Yen, where roughly one dollar is 102 Yen. The prices are comparable to the United States subways such as the Blue Line Metro.

The Sendai Subway was first in operation in 1987. It uses fuzzy logic to accelerate and brake. Only a conductor is needed to ensure safety during operation. The centroid method for use with motion control is used, developed by Hitachi and further expands on the fuzzy logic control method I described above.

The Sendai Subway has two operation modes, train speed regulation control and train stopping control. These operate on two separate but communication fuzzy logic controllers and take a number of calculations into account. These include acceleration to a target speed with velocity oscillation. Maintaining target speed (coasting), and stopping at the correct location with velocity control for a no motion feel. Other factors taken into account are safety, energy consumption, traceability, and passenger comfort.

The Sendai subway has three main areas for subway control and maintenance.

The Administrative Center

There are two departments: the Operations Department which takes care of operations and maintenance, and the Electrical Department which supplies electric power and evaluates the equipment for disaster prevention. They cooperate with each other to maintain safe and efficient operation of the subways.

Train Maintenance Warehouse

Used for Inspection and Maintenance of the Trains. Subway cars are inspected every three days and examined by automatic inspecting devices every three months. They are also overhauled in the factory every three years and six years for two different types of inspections.

Car Base
Subway cars are kept in the safe and good-running condition at the Car Base in Tomizawa. The
Administrative Center, the factory, the train maintenance shed and an automatic car washing machine are all located there.

A second line is planned as an east-west line (Tozai Line): 13.9 km, 13 stations, all underground, and linear motor technology as in Tokyo's O-Edo Line. Construction starting in 2004 for completion in 2016.

Comparisons

I would like to point out a few facts from two similar subways systems. The Blue Line in LA and the Sendai subways system.

Blue Line Facts & figures

  • Length: 22 miles(21.5 miles surface/elevated, 0.5 subway)
  • Groundbreaking: October, 1987
  • Opening: July 14, 1990
  • Cost: $877 Million
  • Ridership 63,000 daily avg (as of 7/00)

Sendai Facts & figures

  • Length: 15 km
  • Groundbreaking: October, 1986
  • Opening: July 14, 1987
  • Cost: $472 Million
  • Ridership 165,000/day (as of 2001)

The future Sendai rail addition is said to cost in the neighborhood of 285 billion yen.

As you can see these two subways are built in similar periods and when correlating the figures they are relatively similar in construction and costs. Note that the blue line is traditional non-fuzzy controllers where as the Sendai subway uses fuzzy logic controllers. Note that the Japanese and USA per-capita income is in less than a 25% deviation.

The Sendia Operational costs per passenger are $0.61 while Los Angeles is around $0.29. This is about double in Sendai what the costs are in LA. The costs do vary in other subway systems however. The skyrail systems in the dollars or even tens of dollars in operational costs per passenger. This is mainly because of the new technology and the high cost on construction. Other rails have a cost as low as $0.15 per passenger, namely in Tokyo/Eidan, although these are known as ‘light rail transit’ and this typically operates in a variety of alignments, including streets, but is capable of operating in trains, and usually includes well-defined passenger stations or transit stops, perhaps with prepaid or other more specialized types of fare collection.

Delineating the Pros and Cons

Sendai is thought of as a success story for fuzzy logic so it is a good example to work with. With this in mind, the other factors in Sendai are comparable to subways build in the same time frame, such as the Blue Line LA rail mentioned above. As the data points out, the operational and maintain ace costs are roughly double for the fuzzy logic system than for the LA Blue Line system. This data is based on per passenger per mile traveled, and both rails are fairly similar with the exception of geography.

Fuzzy Logic may be in everyday use including micro controllers and expert systems, but those are not in particular concern. What is of concern is the efficiency of the Sendai project and no evidence of exceptional efficiency is presented nor have I found any conclusive evidence. Oddly enough I have read about significant savings in energy but I have yet to read exactly where these saving are from. From what I have seen, the time travel is roughly the same, the passenger load is the same, and the distance traveled is a constant. If this is true, then where are the savings? One reference I did read is the idea of coasting when acceleration is not needed. Although this may not be applicable in conventional controllers it seems unlikely as a source of significant savings.

Although the material I reviewed used Sendai in their examples, they are lacking some of specific details of the Sendai subway – specifically how well it is running today. What I did get was a bit of philosophical arguments and I can see the occasional loss of objectivity in the material I have reviewed. If you are pro or con fuzzy logic, it does not matter. This paper is only concerned with the bottom line, and that is I see no significant savings. If anything the Sendai costs are significantly higher than those run by the conventional subway in LA. If you include per person income then the costs are roughly 23% more to operate and maintain the subway. There may be other factors than fuzzy logic naturally – but these are countless and can be swayed one way or the other depending on the argument. The bottom line is Sendai costs more than the Blue Line.

Maglev trains, which do use fuzzy logic as well, have an enormous cost per passenger compared to Sendai. This can be as much as ten times the conventional costs. This likely has nothing to do with fuzzy logic however and is due to the new technological costs and the scope of those projects in the Germany and Japan. The Maglev trains would have probably been a better example to use for a fuzzy logic paper since the data of energy consumption is readily available and a little engineering work can give us some idea on how much energy is being spent if a conventional engineering method was used in its stead.

Conclusion

The use of fuzzy logic is an interesting and useful method for solving AI problems in today’s machinery. In the instance of Sendai, it seems to be in its infancy and it proves that Fuzzy Logic can and does work on a daily basis. While the operational and maintenance costs of the Sendai system are more, fuzzy logic is likely not a direct factor in this. With this same frame of mind, I did not see any conclusive evidence that Fuzzy Logic contributed to any energy savings either. Some of the documentation I have said that this was indeed the case but I have yet to read any data that backs up these claims.

From a wider perspective Fuzzy Logic itself can be seen as a success. Any working operational model should be deemed a success and this is particularly true of fuzzy logic since it solves its problem by very abstract means with application beyond machinery control. Fuzzy logic allows for human like prediction and it likely will have an impact in robotics, market and weather predictions, AI and other fields of study. It is also true that fuzzy logic does indeed provide a better riding experience on the maglev and Sendai rails because of the sense of no motion. If this uses a little or less energy may be irrelevant but a more comfortable ride has other effects that may be more subtle. For example, if the ride is less prone to velocity adjustments that passengers feel, some safety equipment may be irrelevant. It may possible to serve food and other delicate activities that were not possible before. Movement sensitive devices such as networking and computer equipment maybe more reasonable as well. This is in addition to the creature comforts that we as people enjoy.

I have the feeling that future fuzzy logic application may be subtle things and not necessarily any earth shattering all-inclusive methods to solve problems. Fuzzy logic will become another tool to be used in a repertoire of utilities to be used by future engineers. While entire trains and complex systems may not be entirely run with fuzzy logic, it should be very useful for specific tasks. In particular tasks that involve a low amount of precise data or work that needs predictions in a human-like method of solving problems. Fuzzy logic will not replace all conventional machine controllers any time soon however. Likely there will be a need for both with emphasis on the conventional types of controllers.


References

The Logic of Fuzziness, Nazim Soylemezogolu, (http://www.math.harvard.edu/~hmb/issue2.1/FUZZY/fuzzy.html)

Fuzzy Logic: The revolutionary computer technology that is changing our world
by Dan Mcneill
Publisher: Simon & Schuster; Reprint edition (April 14, 1994)

Fuzzy Logic with Engineering Applications
by Tim Ross Publisher: John Wiley & Sons; 2 edition (August 20, 2004)

The Importance of Being Fuzzy
by Arturo Sangalli Publisher: Princeton University Press; (November 2, 1998)

The Fuzzy Systems Handbook: A Practitioner's Guide to Building, Using, & Maintaining Fuzzy Systems by Earl Cox Publisher: Morgan Kaufmann Publishers; 2nd Bk&Cdr edition (January 15, 1999)

Fuzzy Future by Bart Kosko (1999). The Fuzzy Future. From Society and Science to Heaven in a Chip. Harmony Books. New York

Fuzzy Logic for just ‘Plain Folks’ http://www.fuzzy-logic.com/

by Thomas Sowell

Monorail Operational Costs data - http://www.lightrailnow.org/facts/fa_monorail003.htm

Blue Line Metro home site - http://www.westworld.com/~elson/larail/blue.html

Report by:

Samuel E Kisko of West Virginia University

Email: canecorpus@hotmail.com

Back to top