ANEXPERTSYSTEMTOADVISEONURBANPUBLICTRANSPORTTECHNOLOGIESABSTRACT. An important feature of the "sustainable city"concept is a requirement to provideattractive public transport alternatives to the car. In order to achieve this, some cities will require anew public transport system. However, there is no systematic method of deciding which publictransport technology is the most appropriate in a given situation. Expertise in this area is limited tothose who have implemented a new public transport system in their own city, and such expertise is notreadily available to those embarking on the process. The aim of the UTOPIA (Urban TransportOperations and Planning using Intelligent Analysis) project is to develop an expert system to assistwith the decision making process in cities investigating the development of a new public transportsystem. This paper outlines the features required of an expert system, then it goes on to discuss theknowledge acquisition techniques used to elicit information about fourteen British and thirty-twonon-British systems which are either planned or became operational after 1976. Analysis of theinformation gives general knowledge about the nature of the cities, the systems, and the detailed rulesused inthedecisionmakingprocess.INTRODUCTIONCar ownership is increasing rapidly, leading to more road congestion, more use of energy andmore environmental damage.There is much concern about theneed tomake cities"sustainable",thatis, to adopt strategies that are not detrimental to future generations. A key feature of such strategies isthe need to provide attractive alternatives to the car. In many cities, public transport has been allowedto degenerate through lack of adequate investment. It is recognized that in some cities there is a needto provide a new public transport system. There are a variety of possible systems: suburban rail, metro,light rail, improved bus services, busways, guided buses and so on, or some combination of these. Arange of such systems exists in cities around the world. However, there is no systematic way ofdeciding which technology is the most appropriate. Hence there is a danger that each decision will bemade from first principles. In fact, many of those involved in such decisions make only one suchdecision during their life. Therefore, there is a need to transfer experience between one city andanother. Oneway is to encapsulate knowledge about the decision-making process in one city, and thentoapply itelsewhere
AN EXPERT SYSTEM TO ADVISE ON URBAN PUBLIC TRANSPORT TECHNOLOGIES ABSTRACT. An important feature of the "sustainable city" concept is a requirement to provide attractive public transport alternatives to the car. In order to achieve this, some cities will require a new public transport system. However, there is no systematic method of deciding which public transport technology is the most appropriate in a given situation. Expertise in this area is limited to those who have implemented a new public transport system in their own city, and such expertise is not readily available to those embarking on the process. The aim of the UTOPIA (Urban Transport Operations and Planning using Intelligent Analysis) project is to develop an expert system to assist with the decision making process in cities investigating the development of a new public transport system. This paper outlines the features required of an expert system, then it goes on to discuss the knowledge acquisition techniques used to elicit information about fourteen British and thirty-two non-British systems which are either planned or became operational after 1976. Analysis of the information gives general knowledge about the nature of the cities, the systems, and the detailed rules used in the decision making process. INTRODUCTION Car ownership is increasing rapidly, leading to more road congestion, more use of energy and more environmental damage. There is much concern about the need to make cities "sustainable", that is, to adopt strategies that are not detrimental to future generations. A key feature of such strategies is the need to provide attractive alternatives to the car. In many cities, public transport has been allowed to degenerate through lack of adequate investment. It is recognized that in some cities there is a need to provide a new public transport system. There are a variety of possible systems: suburban rail, metro, light rail, improved bus services, busways, guided buses and so on, or some combination of these. A range of such systems exists in cities around the world. However, there is no systematic way of deciding which technology is the most appropriate. Hence there is a danger that each decision will be made from first principles. In fact, many of those involved in such decisions make only one such decision during their life. Therefore, there is a need to transfer experience between one city and another. Oneway is to encapsulate knowledge about the decision-making process in one city, and then to apply it elsewhere
A project to do this is being carried out at University College London, in the University ofLondon Centre for Transport Studies. The project, named UTOPIA (Urban Transport Operations andPlanning using Intelligent Analysis), has the following objectives:· to develop an understanding of how decisions are made about the most appropriate publictransport system for an urban area,?to encapsulate that knowledge in an expert system;totransfer thatknowledge elsewhere?In the next section the nature of expert systems and why they are an appropriate methodology arediscussed. Then the approach being used in the project is described. A method of constructing theexpert system is then presented. Finally, further work is identified and conclusions are drawnEXPERTSYSTEMSExpert systems have been defined as "sophisticated computer programs that manipulateknowledge to solve problems efficiently and effectively in a narrow problem area" (Waterman, 1985)Assuch,theyareusedtosolveproblemswhichusuallyrequirehumanexpertiseorknowledge,forexample, planning, diagnosis and interpretation. It is the explicit inclusion of human knowledge in theexpert systemwhich distinguishes it from othertypes ofcomputer program.An expert system will include two types of knowledge:factual knowledge and heuristicknowledge (including "rules of thumb" and problem solving strategies) (Hayes-Roth, Waterman, &Lenat, 1983). Both types of knowledge are required to solve a problem. While factual knowledgeabout the domain can usuallybe obtained from conventional sources (for example,publications),heuristic knowledge must be obtained from a human expert. The process by which this is achieved isknown as "knowledge acquisition" The method of knowledge acquisition used in the UTOPIA projectisdiscussed inthefollowingsection.An expert system solution is not appropriate for all types of problem. Indeed, Waterman (1985)states that expert system development should only be attempted if it is "possible, justified andappropriate".Waterman's rules for determining when these criteria are satisfied are given in Figure 1.It was felt that these rules actually do apply to the domain of urban public transport planning. Inparticular:1.experts exist (public transport systems have been built) and preliminary investigations showed thatthey can articulate their approach to the problem;2. experts are scarce (relatively few systems have been built or authorized) but there are many
A project to do this is being carried out at University College London, in the University of London Centre for Transport Studies. The project, named UTOPIA (Urban Transport Operations and Planning using Intelligent Analysis), has the following objectives: to develop an understanding of how decisions are made about the most appropriate public transport system for an urban area; to encapsulate that knowledge in an expert system; to transfer that knowledge elsewhere. In the next section the nature of expert systems and why they are an appropriate methodology are discussed. Then the approach being used in the project is described. A method of constructing the expert system is then presented. Finally, further work is identified and conclusions are drawn. EXPERT SYSTEMS Expert systems have been defined as "sophisticated computer programs that manipulate knowledge to solve problems efficiently and effectively in a narrow problem area" (Waterman, 1985). As such, they are used to solve problems which usually require human expertise or knowledge, for example, planning, diagnosis and interpretation. It is the explicit inclusion of human knowledge in the expert system which distinguishes it from other types of computer program. An expert system will include two types of knowledge: factual knowledge and heuristic knowledge (including "rules of thumb" and problem solving strategies) (Hayes-Roth, Waterman, & Lenat, 1983). Both types of knowledge are required to solve a problem. While factual knowledge about the domain can usually be obtained from conventional sources (for example, publications), heuristic knowledge must be obtained from a human expert. The process by which this is achieved is known as "knowledge acquisition". The method of knowledge acquisition used in the UTOPIA project is discussed in the following section. An expert system solution is not appropriate for all types of problem. Indeed, Waterman (1985) states that expert system development should only be attempted if it is "possible, justified and appropriate". Waterman's rules for determining when these criteria are satisfied are given in Figure 1. It was felt that these rules actually do apply to the domain of urban public transport planning. In particular: 1. experts exist (public transport systems have been built) and preliminary investigations showed that they can articulate their approach to the problem; 2. experts are scarce (relatively few systems have been built or authorized) but there are many
proposed systems;3. the planning process cannot be accomplished solely by using conventional mathematicalmodels--experts take qualitative criteria into account and apply heuristic rules; andIFtask does notrequire commonsense ANDtask requires only cognitive skills ANDexpertscanarticulstetheirmethodsANDgemuineexperts existANDexperts agree on sohutions ANDtask is not too difficult ANDtask is not poorly understoodTHEN expert system approach possibleIFtask sohution has a high pay-off ORhuman expertise is being lost ORhuman expertise is scarce ORexpertise needed im manylocations ORexpertise needed in hostile environmentTHEN expert system development justifiedIFtaskrequiressymbolmanipulationANDtask requires heuristic sohutions ANDtask is nottoo easy ANDtask has practical vahue ANDtask is ofmanageable sizeTHENexpertsystemdevelopmentappropriateFIGURE 1. Rules used to determine if expert system development is possible, justified andappropriate (from Waterman, 1985)4. the task has practical value, in terms of improved public transport, and has a high payoff (thedevelopment of a new public transport system is expensive, but an appropriate system will benefit acitysignificantly)However, it is not possible to establish if the development of an expert system is actuallyfeasible untila complete understanding of the decision-making process has been obtained from the knowledgeacquisition phase of the project.If expert system development is appropriate for a particular problem, the resulting system wouldnormallydisplaythefollowingfeatures.· Auser interface. A typical expert system will obtain information from the user, via a question-answerdialogue, about the current problem or scenario. Expert systems have been (somewhat ambitiously)likened to"an expert on the end of the 'phone".In some cases the user will be allowed to volunteer
proposed systems; 3. the planning process cannot be accomplished solely by using conventional mathematical models-experts take qualitative criteria into account and apply heuristic rules; and FIGURE 1. Rules used to determine if expert system development is possible, justified and appropriate (from Waterman, 1985). 4. the task has practical value, in terms of improved public transport, and has a high payoff (the development of a new public transport system is expensive, but an appropriate system will benefit a city significantly). However, it is not possible to establish if the development of an expert system is actually feasible until a complete understanding of the decision-making process has been obtained from the knowledge acquisition phase of the project. If expert system development is appropriate for a particular problem, the resulting system would normally display the following features. • A user interface. A typical expert system will obtain information from the user, via a question-answer dialogue, about the current problem or scenario. Expert systems have been (somewhat ambitiously) likened to "an expert on the end of the 'phone". In some cases the user will be allowed to volunteer
information about the problem before the system requests information. Clearly the user's answers willhave a restricted format (such as yes/no answers, numerical values or selection from a menu)Typically, the users can indicate the degree of confidence they have in an answer.·Explanation.One of the hallmarks of human experts is their ability to justify the conclusion reached,or to explain why a particular item of information is needed. Therefore, for an expert system to becredible, it must be able to give plausible explanations for its decisions. In addition, during the initialinformation-gathering dialogue, the expert system should be able to explain why a particular item ofinformation is requested.?Graceful degradation. An expert system should show graceful degradation, that is, if the informationgiven is uncertain, incomplete or contradicts the knowledge it has about the domain, the expert systemshould produce a less confident answer rather than none at all. The expert system should also be ableto identify when a problem lies outside its sphere of expertise.From the system developer's point of view, an expert system consists of four main components..The knowledge base contains both factual, and heuristic, domain-specific knowledge. Suchknowledge is usually represented using rules, but frames may be used to represent hierarchicalknowledge. Rules are generally of the form:IF Condition AND Condition... THEN ConclusionA conclusion can be deduced if all the rule conditions are satisfied. (FigureI consists of rules definingwhen expert system development is appropriate)..The working memory or global database contains the factual knowledge supplied by the user aboutthe current problem, along with any deductions which have been made based on that knowledge..The inference engine uses the information in the knowledge base and the working memory to solvethe problem. It has two roles: first, selecting which rules to investigate (the strategy), and second,testing if the selected rule is satisfied, and what conclusion can be drawn. If a rule cannot be satisfiedexactly, or some of the rule conditions are uncertain or unknown then rather than the system failing, atentative or uncertain conclusion should be drawn..The explanation system, which was discussed above.One important, but regrettable, feature of expert systems is that they make mistakes. Expert systemsaddress problems for which a mathematical or algorithmic approach is not appropriate. An expertsystem relies on human expertise and since experts are fallible, expert systems will also be. The mostimportant component of an expert system is its knowledge base. The quality of the knowledge within
information about the problem before the system requests information. Clearly the user's answers will have a restricted format (such as yes/no answers, numerical values or selection from a menu). Typically, the users can indicate the degree of confidence they have in an answer. • Explanation. One of the hallmarks of human experts is their ability to justify the conclusion reached, or to explain why a particular item of information is needed. Therefore, for an expert system to be credible, it must be able to give plausible explanations for its decisions. In addition, during the initial information-gathering dialogue, the expert system should be able to explain why a particular item of information is requested. •Graceful degradation. An expert system should show graceful degradation, that is, if the information given is uncertain, incomplete or contradicts the knowledge it has about the domain, the expert system should produce a less confident answer rather than none at all. The expert system should also be able to identify when a problem lies outside its sphere of expertise. From the system developer's point of view, an expert system consists of four main components. •The knowledge base contains both factual, and heuristic, domain-specific knowledge. Such knowledge is usually represented using rules, but frames may be used to represent hierarchical knowledge. Rules are generally of the form: IF Condition AND Condition. THEN Conclusion A conclusion can be deduced if all the rule conditions are satisfied. (Figure 1 consists of rules defining when expert system development is appropriate). •The working memory or global database contains the factual knowledge supplied by the user about the current problem, along with any deductions which have been made based on that knowledge. •The inference engine uses the information in the knowledge base and the working memory to solve the problem. It has two roles: first, selecting which rules to investigate (the strategy), and second, testing if the selected rule is satisfied, and what conclusion can be drawn. If a rule cannot be satisfied exactly, or some of the rule conditions are uncertain or unknown then rather than the system failing, a tentative or uncertain conclusion should be drawn. •The explanation system, which was discussed above. One important, but regrettable, feature of expert systems is that they make mistakes. Expert systems address problems for which a mathematical or algorithmic approach is not appropriate. An expert system relies on human expertise and since experts are fallible, expert systems will also be. The most important component of an expert system is its knowledge base. The quality of the knowledge within
the knowledge base will largely determine the reliability and the robustness of the expert system. Thenext section looks at how the knowledge required for the UTOPIA expert system was obtainedTHEAPPROACHNew public transport systems in both Britain and abroad are being examined. For the systems inBritaintheapproachusuallyadoptedisasfollows:1. Draw up a questionnaire defining areas of interest;2. Identify suitable experts;3. With each expert:Arrange an interview?Send the questionnaire· Interview the expert using the questionnaire as a basis for discussion (the interview is tape recorded).Transcribe the tape recording of the interview, supplementing it with contemporaneous written notes·Analyse the information toderive strategicknowledge and rules;4. Incorporate the knowledge into the expert system;5. Validate the expert system.If an interview cannot take place (usually the case for systems outside Britain) a written version of thequestionnaire is sent.As this tends toreceive a briefer response than a verbal interview, and since notranscription is necessary, it is much quicker to analyse.A number of topics are included in the questionnaire:1. Objectives of building the system,2. The alternatives considered, both in terms of other modes and other ways of achieving theobjectives, such as road building,3. The factors taken into account when deciding on the most appropriate mode;4. Location:.On the surface but away from the highway· Along the edge or median of the highway, segregated from the highway, except at junctions. On the highway, competing with cars for road space· Use of tunnels or elevated structures;5. The nature of the traffic priority re,me if there are conflicts with road traffic at junctions,6. The effects of utilities, that is gas, electricity, water and telecommunications on the constructionprogramme;
the knowledge base will largely determine the reliability and the robustness of the expert system. The next section looks at how the knowledge required for the UTOPIA expert system was obtained. THE APPROACH New public transport systems in both Britain and abroad are being examined. For the systems in Britain the approach usually adopted is as follows: 1. Draw up a questionnaire defining areas of interest; 2. Identify suitable experts; 3. With each expert: Arrange an interview •Send the questionnaire • Interview the expert using the questionnaire as a basis for discussion (the interview is tape recorded) •Transcribe the tape recording of the interview, supplementing it with contemporaneous written notes •Analyse the information to derive strategic knowledge and rules; 4. Incorporate the knowledge into the expert system; 5. Validate the expert system. If an interview cannot take place (usually the case for systems outside Britain) a written version of the questionnaire is sent. As this tends to receive a briefer response than a verbal interview, and since no transcription is necessary, it is much quicker to analyse. A number of topics are included in the questionnaire: 1. Objectives of building the system; 2. The alternatives considered, both in terms of other modes and other ways of achieving the objectives, such as road building; 3. The factors taken into account when deciding on the most appropriate mode; 4. Location: •On the surface but away from the highway • Along the edge or median of the highway, segregated from the highway, except at junctions • On the highway, competing with cars for road space • Use of tunnels or elevated structures; 5. The nature of the traffic priority re,me if there are conflicts with road traffic at junctions; 6. The effects of utilities, that is gas, electricity, water and telecommunications on the construction programme;