a planning methodology for intelligent urban transportation systems

19
Pergamon Transpn. Res.-C. Vol. 2. No. 4. pp. 197-215. 1994 Copyright 0 1994 Elxvicr Science Ltd Printedin Great Britain.All tights reserved 0968-090X/94 $6.00 + .oO 0968-090X(94)00008-5 A PLANNING METHODOLOGY FOR INTELLIGENT URBAN TRANSPORTATION SYSTEMS ADIB KANAFANI, ASAD KHATTAK, and JOY DAHLGREN Institute of Transportation Studies, University of California at Berkeley, 109 McLaughlin Hall, Berkeley, CA 94720, U.S.A. (Received 15 February 1994; in revised form I August 1994) Abstract-Recent developments in intelligent transportation systems pose new challenges and oppor- tunities for urban transportation planning. To meet these challenges and to exploit these opportunities, a framework for a new transportation planning methodology has been developed. The methodology operates in a computer environment, called PLANiTS (Planning and Analysis Integration for Intelligent Transportation Systems), designed to facilitate the entire planning process from problem identification, through idea generation and analysis, on to prioritization and programming. To assist in problem identification, PLANiTS provides graphic representation of current conditions, including traffic, air pollution, accidents, and projections of future conditions. A computerized knowledge base, containing information about possible strategies and their effects, and a model base, containing transportation and other analysis models, are used to guide the user in identifying potentially effective strategies and performing the appropriate analysis. To facilitate the use of these tools, PLANiTS provides computer support of group processes such as brainstorming, deliberation, and consensus seeking. PLANiTS is designed for use in urban transpottation planning at the local, regional, and state levels; it is intended to support a variety of participants in the planning process including transportation professionals, decision makers in transportation agencies (often local elected officials), citizens, and interest groups. Recog- nizing that transportation planning is essentially a deliberative, political process, PLANiTS is designed to inform and facilitate, but not replace, the political decision-making process. 1. INTRODUCTION Urban transportation has been the subject of renewed interest and increased attention in recent years. The continued growth in scale and complexity of urban transportation activities and their impacts on the urban environment has renewed the decades long search for solutions to the urban transportation problem. One important contemporary element of this search is what we might call the technology revival. The last few years have witnessed a rapid increase in the development of transportation technology. The search for solutions has extended to the bound- aries of today’s scientific and technological know-how and has included, most notably, highway automation and the use of information and computer technology for the management of trans- portation systems. The emergence of these technologies as a promising new pathway for transportation is probably the most significant development in the field since the introduction of urban freeways. These developments have resulted in the launching of what has come to be known as the intelligent vehicle highway systems (IVHS). Broadly defined, IVHS implies a transportation system that functions with real-time feedback. Information technology would be used in combination with modem communications and computation technologies to advance the state-of-the-art in transportation systems management and use. These technologies would also be utilized to optimize a system in which increasing proportions of the functions are automated. The anticipation is that IVHS technologies will improve productivity, enhance safety, and reduce the adverse impacts of urban transportation systems. It is also anticipated that new opportunities for innovation in transportation and in the organization of urban socioeconomic activities might be brought about by IVHS. The new developments pose a particular challenge to transportation planning. A central question is whether the processes, methodologies, and tools currently in use are appropriate for planning advanced transportation systems of the types being conceived in the IVHS arena. The system integrating effects of some IVHS technologies, as well as the real-time feedback mech- anisms implied in their operations pose a real challenge to the planning models used to estimate 197

Upload: adib-kanafani

Post on 26-Jun-2016

214 views

Category:

Documents


1 download

TRANSCRIPT

Page 1: A planning methodology for intelligent urban transportation systems

Pergamon

Transpn. Res.-C. Vol. 2. No. 4. pp. 197-215. 1994 Copyright 0 1994 Elxvicr Science Ltd

Printed in Great Britain. All tights reserved 0968-090X/94 $6.00 + .oO

0968-090X(94)00008-5

A PLANNING METHODOLOGY FOR INTELLIGENT URBAN TRANSPORTATION SYSTEMS

ADIB KANAFANI, ASAD KHATTAK, and JOY DAHLGREN Institute of Transportation Studies, University of California at Berkeley, 109 McLaughlin Hall,

Berkeley, CA 94720, U.S.A.

(Received 15 February 1994; in revised form I August 1994)

Abstract-Recent developments in intelligent transportation systems pose new challenges and oppor- tunities for urban transportation planning. To meet these challenges and to exploit these opportunities, a framework for a new transportation planning methodology has been developed. The methodology operates in a computer environment, called PLANiTS (Planning and Analysis Integration for Intelligent Transportation Systems), designed to facilitate the entire planning process from problem identification, through idea generation and analysis, on to prioritization and programming. To assist in problem identification, PLANiTS provides graphic representation of current conditions, including traffic, air pollution, accidents, and projections of future conditions. A computerized knowledge base, containing information about possible strategies and their effects, and a model base, containing transportation and other analysis models, are used to guide the user in identifying potentially effective strategies and performing the appropriate analysis. To facilitate the use of these tools, PLANiTS provides computer support of group processes such as brainstorming, deliberation, and consensus seeking. PLANiTS is designed for use in urban transpottation planning at the local, regional, and state levels; it is intended to support a variety of participants in the planning process including transportation professionals, decision makers in transportation agencies (often local elected officials), citizens, and interest groups. Recog- nizing that transportation planning is essentially a deliberative, political process, PLANiTS is designed to inform and facilitate, but not replace, the political decision-making process.

1. INTRODUCTION

Urban transportation has been the subject of renewed interest and increased attention in recent years. The continued growth in scale and complexity of urban transportation activities and their impacts on the urban environment has renewed the decades long search for solutions to the urban transportation problem. One important contemporary element of this search is what we might call the technology revival. The last few years have witnessed a rapid increase in the development of transportation technology. The search for solutions has extended to the bound- aries of today’s scientific and technological know-how and has included, most notably, highway automation and the use of information and computer technology for the management of trans- portation systems. The emergence of these technologies as a promising new pathway for transportation is probably the most significant development in the field since the introduction of urban freeways. These developments have resulted in the launching of what has come to be known as the intelligent vehicle highway systems (IVHS). Broadly defined, IVHS implies a transportation system that functions with real-time feedback. Information technology would be used in combination with modem communications and computation technologies to advance the state-of-the-art in transportation systems management and use. These technologies would also be utilized to optimize a system in which increasing proportions of the functions are automated. The anticipation is that IVHS technologies will improve productivity, enhance safety, and reduce the adverse impacts of urban transportation systems. It is also anticipated that new opportunities for innovation in transportation and in the organization of urban socioeconomic activities might be brought about by IVHS.

The new developments pose a particular challenge to transportation planning. A central question is whether the processes, methodologies, and tools currently in use are appropriate for planning advanced transportation systems of the types being conceived in the IVHS arena. The system integrating effects of some IVHS technologies, as well as the real-time feedback mech- anisms implied in their operations pose a real challenge to the planning models used to estimate

197

Page 2: A planning methodology for intelligent urban transportation systems

198 A. KANAFANI er al

impacts and assess effectiveness. These same features pose a challenge to the planning process itself, particularly at the local level. Many of the IVHS technologies envisaged require far more coordination in programming and in operations among local communities than is currently the

case. Advanced methods of system management-including options such as automated, dif- ferentiated pricing systems-will raise important questions of regional versus local optimiza-

tion. All these issues will require a planning methodology that is as inrefligent as the transpor- tation systems it is intended to guide.

The need for a new planning paradigm goes beyond the concern with evaluating the impact of new technology. Indeed, other recent developments, which themselves are perhaps not unrelated to the emergence of IVHS, have made it imperative that such work be undertaken. Prominent among these is recent legislation, such as ISTEA (Inter-modal Surface Transportation Efficiency Act) and CAA (Clean Air Act), which mandate planning processes at various levels including urban, regional, and state. There are also cogent reasons for seeking new techniques and processes for transportation planning. Current methods rarely satisfy the needs of the planning and programming processes that have become a part of urban governance. Models are rarely used effectively in informing these processes, and the gap between the methods re- searcher and the policy analyst remains fairly wide.

On the positive side, there have been significant developments in recent years in computer- aided planning and decision support systems. Computer-based support systems have been developed for complex deliberative and negotiating processes of the kind that has become common practice in transportation planning. These systems have typically been applied to private sector problems, but they hold much promise for application to transportation. Decision support systems have also been developed and successfully applied to complex, multiobjective planning processes in many fields including environmental planning (Guariso and Werthner, 1989), large-scale public works, and recently to transportation. These methods hold much promise for urban transportation planning in the IVHS era. One of the goals of this work is to explore how such techniques might be developed for urban transportation planning.

This paper describes our initial efforts to develop such a methodology. Full development is an ambitious undertaking that will take many years. In fact, many of the tools needed for its development are still in early stages of their development. Furthermore, the body of transpor- tation research on which the methodology is based needs further organization and progress.

2. IVHS-IMPLICATIONS FOR A NEW PLANNING METHODOLOGY

To define a planning framework for the IVHS era, it is necessary to adopt a meaningful definition of IVHS. This is a complex subject and one that is evolving as IVHS gains momen- tum in the research and development community, as well as among the agencies that develop and implement transportation policy and programs. There is little disagreement that IVHS represents a stage in a continuing search for improvements to transportation technology. As such, IVHS can be said to include all that is new in the way transportation systems are designed, implemented, and managed. Indeed, in many respects it is difficult to find the difference between some of the technologies referred to as IVHS and their predecessors. For example, much of what is now considered advanced transportation management systems (ATMS) is a continuation of a long tradition of traffic management improvements including the development of computer models such as FREQ and TRANSYT and encompassing the popular transportation system management (TSM) of the 1970s.

Two fundamental features set IVHS apart from earlier developments. One is the use of real-time information and feedback for the management and operation of the system; and the other is the introduction of automation. The first involves the use of advanced communications, computation, and information technologies to permit users and operators of transportation systems to optimize various aspects of the system. Traffic managers can use these technologies to integrate management and control, particularly between freeway and city street subsystems; travelers can use real-time information to better integrate trip-making decisions into the complex of urban activity scheduling, resulting in better use of system resources, and possibly at a better level of service. The introduction of automation also changes transportation systems in a fundamental way. Beginning with driver aids such as collision avoidance and hazard warning,

Page 3: A planning methodology for intelligent urban transportation systems

Planning methodology for urban transportation systems 199

these technologies can improve safety and efficiency, possibly resulting in improved traffic streams with fewer incidents and reduced adverse environmental impacts. Moving on to more extensive applications would yield automated transportation systems where significant capacity

gains are added to the safety and environmental impact gains.

In addition to direct system gains, it is normal to expect that these new features of transportation systems will inspire off-system, higher order gains that can far outweigh them. The introduction of advanced information technology into the use and management of trans- portation systems could inspire fundamental changes in the way urban activities are conducted, creating opportunities for doing things in ways not yet imagined. Likewise, the introduction of automation may spawn innovations in the design and manufacturing of automobilies and their propulsion systems. Perhaps the most important impact of IVHS technologies is their role in catalyzing innovation in the way we do things and the way we use transportation to do them.

All this represents challenges and opportunities to transportation planning. In order to support decision making regarding intelligent transportation systems, the planning process itself must be intelligent. If the transportation system is to have real-time feedback in its operations, then the planning process must include models that reflect that feedback. Information available in the IVHS environment for system operations and management should also be available for system analysis and planning. The continuous feedback in IVHS operation should be echoed by continuous forecasting in the IVHS planning models. The introduction of new elements of information technology, communications, and automation into transportation systems should be reflected in the way the transportation options are conceived and analyzed in the planning process. The wealth of technological options that will become available in transportation sug- gests that the planning process should be capable of dealing with a conrinuum of options rather than a discrete set of alternatives. It should be capable of efficiently searching through this continuum and matching them against policies and objectives to ensure that no opportunities are missed. Finally, all this added complexity must be somehow integrated into a complex decision- making process, one that seems to reflect apprehension about new technology and ambivalence

about modeling and systems analysis. IVHS presents opportunities for enhancing the transportation planning process itself in a

significant way, and the proposed methodology aims to take advantage of these. As real-time operations monitoring is implemented, there will be an explosion in the data available for analysis. Current models designed to use minima1 information will give way to models and knowledge bases that can organize and use rich bodies of data. The new technological options that will become available suggest that the planning process must be capable of dealing with a near continuum of alternatives and should allow for efficient searching through this continuum to match options against policies and objectives.

These challenges and opportunities suggest a planning process that can take advantage of the wealth of information available in IVHS and that has at its disposal the analytical power to use this information intelligently. The process we propose aims to integrate planning and analysis and to provide a computer-based platform within which simple and complex analysis and deliberation are supported. The basic principle of the proposed planning framework is the intelligent use of knowledge to support deliberation and decision making. To operationalize this principle we introduce two important features of the planning framework. The first is to deal with transportation planning as a deliberative, dialectical process that seeks agreement on programming decisions. The second is to supplement models with expertise and with a knowl- edge base that is enriched as experience with new technology is gained. The methodology proposed to implement these principles is computer-based and uses an interactive online envi- ronment to facilitate deliberation and to integrate it with analysis.

2. I Transportation planning as deliberation Transportation planning is primarily a deliberative process of negotiation and consensus

building that is supported, rather than driven, by analyses and projections. The contemporary context of transportation planning is one in which there is a diverse group of actors and stakeholders who are driven by different motives and are advocating conflicting objectives; who have different value systems with which they measure their expectations from the transportation system and with which they judge its impacts; and who are all vying for a common, usually

Page 4: A planning methodology for intelligent urban transportation systems

200 A. KANAFANI er al

limited, resource pool. It is a context of dialectical tension between opposing forces. Recent legislation has made it mandatory that planning be multifaceted, multimodal, and multiagency. The broadening of the scope of transportation planning and the decentralization of transportation planning powers has brought many actors into the planning and decision-making processes, and has made resolution seeking a central feature of these processes.

To deal with this aspect of planning, we place at the heart of the methodology a computer- based intelligent facilitator and decision support system. Planners work within an interactive

online environment that informs, educates, facilitates deliberation, and assists in synthesizing positions and seeking consensus. This consensus seeking is not limited to the final stages of programming but occurs at all stages of the process. Planners need to consider goals, criteria, constraints, models, and predictions before they can accept the results of analysis and come to a consensus on programming. Of course, the methodology cannot guarantee that consensus will be achieved, but it facilitates the process of seeking consensus. Using its rich knowledge base and powerful analytic tools this computer-based intelligent facilitator seeks to discover win-win propositions; to clarify trade-offs in meaningful, and when possible, quantitative terms; and to support trade-off analysis when optimal solutions are not possible.

2.2 Knowledge base in support of planning It has always been true that models cannot totally replace expertise and human judgment.

While modeling is an essential approach to the analysis of complex systems, it remains inad- equate as an intelligent support base for planning and programming decisions. Recent devel- opments in computer science and in data base management techniques have made it possible to assemble large quantities of data regarding the behavior of complex systems and their environ- ments and to extract from these data bases useful knowledge. Expert systems have been developed and applied in many fields, including transportation, to store knowledge and exper- tise in a way that allows its efficient use within the framework of a computer-based decision support system. Furthermore, in a fully developed IVHS-based urban transportation environ- ment, we would expect a fully connected system with real-time feedback used in its operation and management. Such information systems provide a very valuable resource to measure and monitor behavior and from which to build a knowledge base for planning.

2.3 Computer-supported deliberation and analysis The proposed methodology integrates analysis and decision making in an interactive en-

vironment. This requires a substantial computer-aided decision support system. The computer system includes two main elements. One is the knowledge and methods base. It includes the data bases and a database management system, the knowledge base and the collections of methods, and models and tools that perform analysis. The other element is a computer-based deliberation support system. This is a system that facilitates the sharing of information, ideas, and views as part of the deliberation that takes place in planning and decision making. The computer support of the process permits the search through rich data and knowledge bases and allows the users to explore alternatives from a rich array of technologies and other interventions that reside in what is called the action base.

3. PLANiTS-A SYSTEM FOR INTEGRATED PLANNING AND ANALYSIS

PLANiTS, (Planning and Analysis Integration for Intelligent Transportation Systems) is a comprehensive tool designed to meet the needs of the emerging planning processes of the IVHS environment (a detailed description of PLANiTS appears in Kanafani, Khattak, Crotty, and Dahlgren, 1993). Recognizing of the essentially political nature of transportation planning, it is intended to inform rather than replace political decisions.

PLANiTS provides an integrated computer platform for performing the analysis and mod- eling involved in planning, as well as supporting the deliberative processes involved in making programming decisions. While analysis and modeling have been widely used in transportation planning for many years, the electronic display of traffic information and the effects of various actions, the use of expert systems to guide and perform analysis, and the provision of electronic support for group processes are new components that PLANiTS will contribute to the planning

Page 5: A planning methodology for intelligent urban transportation systems

Planning methodology for urban transportation systems 201

process. Also, new with PLANiTS are expanded opportunities for considering alternative assumptions, goals, and strategies for addressing transportation problems. Finally, PLANiTS can be accessed and used by many agencies and interested citizens for brainstorming, devel-

opment of strategies, diagnosis, dialogue, and testing of programs (Fig. 1).

3.1 PMiTS components The basic components of PLANiTS are shown in Fig. 2 and the details appear in Fig. 3.

The central functions of deliberative planning and analysis, and consensus seeking and project programming are supported by four bases residing in an interactive computer environment. These are:

The Policy and Goals Base, which contains mandates, objectives and constraints communi- cated in terms of appropriate measures of performance. The Strategy and Action Base, which contains a catalogue of possible actions and action affiliation relationships. The Data and Knowledge Base, which contains, or provides access to data bases and contains knowledge displayed in terms of established relationships between transportation objects. The Methods and Tools Base, which contains transportation models and generic methods of analysis, and utilities and tools, e.g. for network analysis.

3.2 Policy and Goals Base Planning and programming deliberations depend on fundamental policies, goals, and ob-

jectives, as well as mandates and funding opportunities. Deliberations are often hampered by the lack of clarity regarding the relative importance of these factors. In the Policy and Goals Base we identify, catalogue, and represent the overall policies, mandates, and requirements of the planning process. Mandates such as those found in congestion management legislation will reside here and be used in the selection of strategies and actions, their analysis, and the deliberations on their programming. Also in this base is information on funding opportunities for various actions intended to meet different policy objectives.

Measures of performance and other criteria for evaluation would also reside in this base. PLANiTS contains an extensive list of measures of performance used in transportation planning that are grouped around goal clusters such as delay, air quality, safety, and accessibility. Users will also be able to enter customized measures suited for specific analyses. Different users may value effects differently. For example, a traffic engineer concerned about traffic backing up into residential neighborhoods may be concerned primarily with delay and noise, whereas an environmentalist might be concerned primarily with overall air quality or with the emissions of specific pollutants.

The policies are represented in textual form and as rules. For example, one of the ISTEA “fifteen-factors” is whether a project entails capital investments that would increase security in

POLICY AND GOALS BASE KNOWLEDGE BASE DIAGNOSTICS

STRATEGY AND METHODS BASE ACTIONS BASE

\ STRATEGIES / ANALYSIS

Fig. I. PLANiTS process structure

Page 6: A planning methodology for intelligent urban transportation systems

202 A. KANAFANI et al.

_---_----_ I I I I I( 31 1 DELIBERATIVE PLANNING I

I AND ANALYSIS

I Smregy and Acuan base T A~,t0ll~-* -

Measures ofPerformance-I’ i Methods and Tool Base

I Enwonmmr-E I

Pobcy and Goals Base CONSENSfS$JILDING I RESOLUTION SEEKING I

I Integrated plannmg

I

’ and analysis process 1 I J ---------

Fig. 2. PLANiTS components.

transit systems. Because project approval depends to a certain extent on compliance with these factors, this information can be valuable to users. Furthermore, it can be operationalized into

rules as follows:

Rule

IF AND

Goal is G = {g,} Policy factor of interest is

P = {Pi)

Simplified example

{improve safety} {ISTEA, Personal security in a transit system}

AND THEN

Environment is E = {ei} Actions to be considered are

A = {ai)

{XYZ transit system} {Television surveillance, station agents}

A complete fact list and inference mechanism will be developed for the Policy and Goals

Base along the same lines.

3.3 Strategy and Action Base

The Strategy and Action Base contains a set of actions to increase transportation supply, reduce demand, increase accessibility or otherwise improve system performance, as well as sets of supportive actions that would enhance the effectiveness of the primary action, and compet- itive actions that tend to diminish its effectiveness. For example, a primary action could be the construction of high-occupancy-vehicle or HOV lanes for which a set of supportive actions will include:

l Construction of park-and-ride facilities l Construction of exclusive HOV ramps and flyovers l Enforcement 0 Advanced public transportation systems (APTS) offering real-time ride-share matching 0 Parking pricing and employer-supplied transit and ride-sharing incentives

A competing action would be construction of a rail system serving the same corridor as the HOV lane. The way actions are combined and implemented is critical in determining their impacts.

For example, the joint implementation of ATIS (Advanced Traveler Information System) and ATMS technologies may provide greater benefits than the sum of their individual benefits.

Page 7: A planning methodology for intelligent urban transportation systems

Planning methodology for urban transportation systems 203

Data Base ,managemen, ywll,

Case Base @c.asonm~)

OperaUOll\

SUPPb

Demand

STRATEGY AND ACTION BASE

DATA AND KNOWLEDGE

BASE

f DELIBERATIVE PLANNING AND

ANALYSIS

(,_ flA.Y.El ,

I

7, .

. ‘m .

. . Gencrlc Models

Transponatmn Models

METHODS AND TOOLS

BASE

POLICY AND GOALS BASE

CONSENSUS BUILDING AND RESOLUTION

Fig. 3. PLANiTS Component details.

Table 1 is an “action-interaction” matrix that shows which actions are mutually synergetic and which ones are at cross purposes. The matrix may also show dependencies among actions. For example, field operational tests may show that advanced traveler information systems cannot be implemented without a comprehensive advanced transportation management and surveillance system. Another example of dependence among actions is that suggested by Varaiya and Shladover (1991), who argue that ATIS and ATMS should be designed to accommodate more futuristic automatic vehicle-control technologies. The action-interaction matrix can be opera- tionalized into rules as illustrated below:

Rule Simplified example

IF Goal is G = {g,} {Improve air quality} AND Policy factor of interest is P = {pi} {ISTEA, Environmental effects of

transportation decisions)

AND Measures of Performance Y = {yi} {Carbon monoxide} AND Environment is E = {ei} {ABC county}

AND Action selected is A = {ai} {Purchase buses} THEN Associated actions to be considered {HOV lanes, Advanced Public

are A, = {ajlil Transportation systems/ Purchase buses}

Actions are defined by their attributes. For example, HOVs on a freeway are defined by the number of HOV lanes, the threshold for vehicle occupancy, length, and the times of their

operation. Analysis may be triggered by the action’s potential for mitigating a particular problem or

because of interest in the action itself. For example, consideration of a light rail system may arise because of the need to reduce congestion in a heavily traveled corridor or because an abandoned railroad right-of-way has become available. Consideration can be triggered in three specific ways:

0 A PLANiTS user may propose the action or group of actions l PLANiTS’s expert system may suggest consideration of a potential action, based on the

environment and goals 0 A proposal may be generated through a brainstorming session within PLANiTS

Actions should be evaluated in terms of their effectiveness in meeting goals relative to their costs. Table 2 presents a taxonomy of actions based on their effects. The system impacts may

Page 8: A planning methodology for intelligent urban transportation systems

204 A. KANAFANI et al.

Table I. The action-interaction matrix (selected examples)

Associated actions

Proposed actions

ATIS .Pre-trip info systems .Enroute info systems

ATMS ~Signal conuol .Ramp metering .Incident mgt.

AVCS .Lateral control .Longitudinal control

Driver warn. & assist Collison warning .Obstacle detection .Smart cards (AVI)

Tele-technologies ~Tcleworking .Teleshopping

HOV IatteslRideshare match Rail Bus transit Congestion pricing

+ +

t t

+ +

+ + + +

+ + + + + + + +

+ + + + + + + f + + + + + + + +

+ + + + +

+ +

+ + + + + + + +

+ +

+

+ +

+ + + + - +

i + + + + +

+ = Positive interactions. + + = Required/mandated actions (future). - = Negative interactions.

be general, measured in terms of performance criteria such as congestion and accessibility, or action-specific, measured in similar metrics yet accounting for unique aspects of the action, e.g. changes in neighborhood traffic due to ATIS. The traveler impacts may be tangible from the perspective of the individual, e.g. travel time savings and increased accessibility, or intangible, e.g. increased comfort and convenience.

Because the impacts of actions are the result of individuals’ choices, we must understand how potential actions would influence these choices. The choices include lifestyle decisions, accessibility decisions, and travel (and trip substitution) decisions. Figure 4 shows that trans- portation improvement actions can be mapped to specific choices. Importantly, the attributes of actions influence choices. For example, the design features of a traveler information system such as whether it is in-vehicle or out-of-vehicle may have different impacts on travel decisions. Individuals are expected to respond to the attributes of the actions. The following paragraphs provide examples that focus on new technologies:

Long-term lifestyle decisions. The decisions to form families and participate in the labor force may be impacted by tele-technologies. For example, a person who previously would have stayed home to raise children may now be able to do so while participating in the labor force through tele-working. Conversely, individuals may be more likely to form families because tele-technologies allow them to work at home. Medium-term accessibility decisions. Automated highways may change people’s residential, work location, and automobile-ownership choices. By decreasing the travel time between home and work, they might allow people to relocate further from their work places and might make automobile ownership even more attractive by lowering the cost of travel. Altema- tively, new transit technologies may make travel by transit or rideshare more appeaiing- decreasing the need for owning automobiles.

Page 9: A planning methodology for intelligent urban transportation systems

Planning methodology for urban transportation systems

Table 2. impacts of selected actions

205

Measures of performance

System Individual

Action- General specific Tangible Intangible

Actions

ATIS ‘Pre-trip info systems .Enroute info systems ATMS .Signal control .Ramp metering .Incident mgt. AVCS .Lateral control .Longitudinal control Driver warn. & assist Collison warning .Obstacle detection Tele-technologies .Teleworking .Teleshopping HOV lanes/Rideshare match Rail Bus transit Congestion pricing

+ + + + + + + + + + - + + +

+ + + + + + - iI- -

+ + + + +

+ + + + + + + + + +

+ +

+ + + + + + + + + + + •t + + + + + + + + + + + + + + +/-

+ +

+ +

+ = Positive impacts - = Negative impacts

Short-term trip substitution decisions. Individuals may substitute teleworking, teleshopping, teleconferencing, and telerecreating for travel. However, the automobile freed by the person working at home may be used by the spouse or a child who would not otherwise have access

to the automobile. Short-term travel decisions. Short-term travel decisions consisting of mode, destination, pre-trip route, en route diversion/return, departure time, parking, trip chaining, and trip frequency decisions may be directly influenced by traveler-information technologies and by early versions of automated systems, such as driver-warning and assistance technologies.

These decisions when aggregated result in total travel or transportation system performance. Overall, the Strategy and Action Base provides a mechanism for searching, synthesizing, and refining sets of relevant actions to explore their impacts.

3.4 Data and Knowledge Base The Data and Knowledge Base is a multilayered warehouse of data at various levels of

information and knowledge content used to support the planning process. Methods of knowl- edge extraction and case-based reasoning are used to convert data into relevant knowledge that is generic (i.e. more generalizable than the data from which it derives). This knowledge is used to support the brainstorming process of system diagnosis and problem definition. It is also an important resource for enriching the systems that advise the user on the selection of models and methods of analysis.

The data and knowledge base has three components: a data base, a case base, and an expert

Page 10: A planning methodology for intelligent urban transportation systems

206 A. KANAFANI etal.

Aggregate pred,cuons of 4

\ys,em performance

LONG-TERM LIFESTYLE DECISIONS

Family formalron Labor force panlclpallo”

i

‘I

‘1 I I’

4 MEDIUM-TERM f ‘\

ACCESSIBILITY DECISIONS I Automauc vehxlc Rcsldent,al lccauanl!ypc Employment locauon

‘;-) comrol technologies I

Auto ownerrhtp

1

SHORT-TERM TRAVEL DECISIONS

Mode Desunauon Dcpanure tmle Pre-rnp rO”te Enroute dwer\,on/return Tnp chammg Tnp frequency Tnp supresslonisubsulullon

4,’ I

I\,’

I’

; ‘\ Advanced transponalmn managemenr and mformauon systems

- CapabWy enhancement rechnologlo

Fig. 4. Influence of technologies on individual choices

base. Initially, knowledge from IVHS field operational tests and early experiments will be represented in the data base and the case base. As a sufficiently large body of knowledge develops, the expertise will be transformed to rule-based expert systems.

The data base will have information on existing transportation facilities and operations, existing and projected land use, and travel patterns and demographics. This information can be accessed through user queries. For example, a user might ask for the current delays incurred by a particular group of travelers on a particular link during the P.M. peak period. (An elaborate data structure for representing actions, performance measures, and the environment is currently under development.) The case base contains a catalog of previous cases. Depending on the action of interest, performance criteria, and the context, appropriate cases are retrieved and displayed. For example, when analyzing the impacts of HOV lanes in a corridor, a user can retrieve other HOV studies from the case base; the cases will be displayed in the order of their similarity to the current case. The expert base is a catalog of expert systems which support the diagnosis of problems and search for solutions (Ritchie, 1987; Tung and Schneider, 1987).

Acceptability of studies for the case and expert bases will be determined through evalua- tions by researchers and practitioners. The evaluators will use the following criteria to decide what material should be included in the case base and the expert base:

0 Relevance to planning decisions 0 Conceptual validity, sophistication, and simplicity 0 Contribution to transportation practice and to the body of knowledge regarding transportation l Generalizability and potential for providing useful relationships and rules 0 Originality

Collecting, integrating, processing, refining, filtering, and structuring knowledge of transpor- tation systems for representation as cases and rules will be done within PLANiTS.

To develop case and expert bases, we must understand the nature of the relationships we might extract from theory and data and how might we use them. Theoretical relationships can be causal or a-causal. A simple example of a causal theoretical relationship is that higher travel time and monetary cost of a transportation mode reduces chances that people will take it. Example of an a-causal relationship is that one technology dominates or replaces another one. Relationships can be represented more formally:

4 = Information in ATIS database = {transit information, traffic delay information, ocean terminal delay information, parking information, weather information, etc.}

23 = Travel options = {take transit, drive alone, bike, walk, etc.}

Page 11: A planning methodology for intelligent urban transportation systems

Planning methodology for urban transponation systems 201

Define Ri(a,P) to be the relationship type i for any pair belonging to sets 4 and 93. Suppose that we define &(o,P) = o. causally increases the chance of l3. Then, R&transit information, transit use) = “transit information increases the chance of taking transit.”

There are three types of relationships:

&<, = Theoretical causal relationship

a__, = Theoretical a-causal relationship

4” = Observed (empirical) relationship (may be theoretical or unexplained)

a_,, = No relationship

If R, (a,P) exists, where (Y E &’ and l3 E B, then,

means that (Y and p are theoretically causally related. For example, information on parking availability might cause an increase in the chance of auto use. Another important set of rela- tionships is a-causal:

means that Q and l3 are theoretically related and the relationship is a-causal. For example, ATIS may replace more rudimentary information technologies.

Relationships can be observed empirically. A simple example of such a relationship is the observation that better quality traffic information increases the chance of auto use. If R,(a,f3) exists then,

means that cr and p are observed to be empirically related (though they may or may not be related theoretically).

R,(cY,~) indicates that there is no relationship between some elements of ~9 and B. This could be written in the following forms:

means that there is no causal relationship between (Y and l3. For example, ocean terminal delay information does not cause a change in the chance of using transit.

means that there is no a-causal relationship between (Y and l3.

means that there is no empirically observed relationship between a and p. For example, observation might indicate that weather information does not increase travelers’ chance of taking public transit. Notice that it is possible to have multiple relationships in more than two variables such as “X increases the chance of .v which decreases the chance of z. *’ The use of &_ ,, means that none of the theoretical causal, theoretical a-causal and empirical relationships exist between LY and j3.

One important use of the above structure comes in providing prescriptions or suggestions that should be based on theoretically and empirically valid results. A simple example is that the actions that pass through the theoretical and empirical relationship filters should be recom- mended for consideration. In recommending associated actions, we must first consider whether the action is theoretically plausible (o a_+ p that is, two actions have positive interactions) and

Page 12: A planning methodology for intelligent urban transportation systems

208 A. KANAFANl et al.

also whether in real-life situations they have been observed to work (a &g B that is, two jointly implemented actions have worked well together). Finally, the solution sets for various rela-

tionships are:

di = {(cu,B) ) a E 4 and B E ZB and R,(x,y) exists} where i is the index for relationship type.

With this structure in mind we now examine each base individually.

3.4.1 Data base The data base is based on interviews/focus groups, cross sectional and longitudinal sur-

veys, vehicle movement logs, and human factors data. It will allow users to obtain information on travel patterns, accessibility decisions, and lifestyle choices. The system performance data, included in PLANiTS, is categorized as genera1 and action-specific. The genera1 category consists of, for example, induction loop detector data, safety data, energy consumption data, and pollution measures. The action-specific category has data on traffic information system performance and control system performance. System performance can be represented on a

geographic information system. However, this does not preclude other data-representation

methods.

3.4.2 Case base Humans acquire, process, and store useful information in their memory and later retrieve

it to address new situations. They learn from their experience and are often able to better address similar future situations. Case-based reasoning (CBR) uses the idea of learning from experience in the context of computers. Experiences stored in the computer memory are recalled to address new situations. CBR relies on simple logic rules; for example, it can recommend actions that worked, and it can warn against those that did not work.

CBR is a relatively new paradigm in artificial intelligence (Kolodner, 1993; CBR, 1989, 1990). It determines the similarity of past cases to a present situation, retrieves relevant cases from computer memory and informs the user about how similar situations were addressed and whether the solutions were successful. Causal models of past cases can be represented to obtain insights and explanations. However, a causal mode1 is not required, which is useful when causality cannot be applied easily, such as in predicting higher order effects or when knowledge is very limited. There is considerable experience with precursors to new technologies that can be used to seed the case base. However, the experience is not sufficient for mode1 building and forma1 inference. Initially, new technology implementation decisions can be made by examin- ing evidence from field operational tests; successful field operational tests can be replicated and past mistakes can be anticipated and avoided.

An important limitation of CBR in transportation planning is that no two cases are exactly similar. Judging similarity of cases is often difficult and sometimes arbitrary, because the methods for comparing cases are not well developed (CBR, 1989, 1990). Further, even in similar cases, it is often difficult to modify the previous cases sufficiently to provide useful insights and predictions.

PLANiTS can use case-based reasoning at the following stages:

0 During the preliminary analysis of a particular action it can inform users about the impacts of similar actions implemented elsewhere

0 During more advanced analysis stages it can compare the relationships between variables and data for the current case with previous cases and determine whether the parameter estimates for a current case are consistent with earlier cases.

In addition to presenting cases to users, PLANiTS can display them along with descriptive and prescriptive, (or both) types of information. An example of descriptive information, when constructing HOV lanes is being considered (based on the previous discussion about relstion- ships), is that “given certain travel patterns, HOV lanes reduce traffic congestion.” Such information about relationships must go through the appropriate theoretical and empirical filters before it is presented to a user. An example of descriptive and prescriptive information is that “given the similarities of travel patterns between corridor X and the corridor under consider- ation, and that HOV lanes along with park-and-ride lots reduce traffic congestion, PLANiTS recommends jointly evaluating HOV lanes and park-and-ride lots.” Further, not only the

Page 13: A planning methodology for intelligent urban transportation systems

Planning methodology for urban transportation systems 209

positive consequences of actions will be presented but also warnings against actions that did not produce the desired impacts or that resulted in unanticipated negative consequences.

The PLANiTS case base relies on the development of a data structure for indexing and storing relevant transportation cases and on the work of researchers in the CBR domain. At the core of case-based reasoning is the nature of comparison between cases. These can be of the following types:

0 Comparison of qualitative criteria and descriptions for the two cases. For example, compar- ison criteria can be whether traffic congestion is a serious problem in the region of interest and whether infrastructure is available to support an action.

0 Comparison of underlying relationships in the two cases. A conceptual model can have theoretical (causal, a-causal) and empirical relationships between concepts and physical entities. Further, class hierarchies may exist between system elements. Whether or not the important relationships and structures hypothesized in the current case are also present in the retrieved case will influence similarity among cases.

0 Comparison of quantitative attributes between the two cases. The quantitative attributes can consist of raw data such as design characteristics for a technology and transportation infra- structure or processed data such as descriptive statistics and parameter estimates.

The cases, collected from the literature and field operational tests, will be indexed according to actions, and performance measures, and their environment. Once a preliminary match is made, similarities to and differences from the present case will be evaluated to decide whether the case is relevant. The user will be given information on the nature of differences between the present and retrieved cases.

In the analysis of complex transportation problems, a previous case may address part of the problem, requiring additional cases to address the remaining problem. For example, to analyze the impacts of HOV lanes on congestion, an equivalent case may be found. However, the same case may not be useful if congestion impacts of combined implementation of HOV lanes and ramp meters were needed. A structure will be developed to organize and use knowledge from different cases to address complex transportation problems.

3.4.3 Expert Base When faced with a problem, individuals have the ability to process information stored in

their memory and reason with the available knowledge. Knowledge-based systems (KBS) mimic the memory and reasoning capabilities of human beings. In so doing, their objective is to equal human problem solving abilities. Such systems are being developed widely in trans- portation (Ritchie, 1987; Tung and Schneider, 1987).

Knowledge-based expert systems (KBES) provide a natural progression from case-based reasoning systems; when sufficient case knowledge is accumulated, it can be transferred into the expert base. KBES can consider imperfect information (e.g. when relationships between system components are not known reliably) and solve mathematically ill-defined problems. Information that can be used by a KBES can be generated from selected literature, human experts as well as from field operational tests, and technology deployment. The main limitation of expert systems in the context of transportation planning is that problems are usually complex and there are seldom “experts” or studies that the participants can mutually agree upon.

Use of an expert system requires a set of inputs that describe the current transportation problem; the inputs are processed through an inference mechanism that has functions and operators and produces a solution/advice. Results are communicated to users through a set of outputs. Consequently, the main components of KBES are knowledge (facts, relationships, and heuristics represented on a continuum of soft and hard rules), an inference mechanism and a user interface. The inference mechanism has a set of rules that can be augmented, changed, or deleted by system developers. The rules have antecedents, which are conditions that should be satisfied, and consequents, which are the actions to be taken when certain conditions are satisfied. The inference mechanism can be data-driven or goal-driven.

Most current KBES are rule-based and often use empirical relationships and/or subjective knowledge acquired by human experts. Such systems do not perform adequately beyond the boundaries of their knowledge domain. The PLANiTS KBES will also embody theoretical

Page 14: A planning methodology for intelligent urban transportation systems

210 A. KANAFAM et al.

models (causal and a-causal relationships). Such systems, when developed, will presumably perform better than systems based solely on empirical relationships. Thus, KBES in PLANiTS will have conceptual models that allow deeper reasoning.

The catalog of expert systems will include diagnostic and action support modules. These help users identify problems and explore solutions. An action support design module for HOV lanes could rely on the following rules.

Theoretically based rules. If the number of vehicles eligible for an HOV lane use exceed the capacity of the HOV lane (1800 vehicles per hour per lane), then an HOV lane will not perform any better than a mixed-flow lane in terms of person delay and vehicle emissions. Empirically based rules. If travel-time savings offered by HOV lanes are greater than a threshold (one minute per mile), then significant mode shift to HOV is expected. If HOV lanes are longer than a threshold (e.g. two miles), then they reduce congestion (except on bridges and bottleneck bypasses where they can be shorter).

The development of a knowledge base for transportation planning represents a major undertaking. Our aim is to capture the knowledge available from the rich data sources that exist

in transportation and from the experience gained through field operational tests and demonstra- tion projects of new technologies. We believe that PLANiTS presents a significant opportunity to advance the state-of-the-art in transportation knowledge representation and management, and to integrate the wealth of existing information used in analyzing transportation systems.

3.5 Methods and Tools Base The Methods and Tools Base is also multi-layered. It contains a methods base with analysis

methods at different levels of intensity and specificity. It provides options varying from aggre- gate simulations to detailed disaggregate models that are more specific in nature. The base also contains a set of tools that provide analysis support through generic statistical and network analysis models. The results of model runs, after they are filtered, are fed back into the knowledge base. Thus with repeated applications of PLANiTS, the knowledge base will grad- ually become richer. In certain situations, the need for detailed modeling and elaborate data collection may decline.

The Methods Base allows users to run existing models or develop new models. It will contain existing transportation models, such as UTPS-type planning models with trip genera- tion, trip distribution, mode choice, and route choice components; and operational models such as FREQ and TRANSYT. To integrate these models, their compatibility will be studied by identifying their inputs and outputs. The context in which the models will be applied is de- scribed by equations and their relevance to specific situations.

Users will be able to run compatible models together, enhancing the ability to evaluate impacts. Depending on the problem being addressed or actions being explored, models from different transportation domains can be joined to conduct more comprehensive analyses. For example, the outputs of a model that analyzes change in travel times due to availability of real-time traveler information can be used as inputs to a model that predicts any subsequent changes in travelers’ selection of various transportation modes.

PLANiTS will provide information on the theoretical structure, validity, accuracy and data requirements, and inputs and outputs of all models and their submodels in the system. This information will be used to develop rules that are structured as follows:

Rule Simplified example

IF AND

AND

AND AND

Goal is G = {gr} Policy factor of interest is

p = {Pi) Measures of Performance

y = {Yi) Environment is E = {ei} Action selected is A = {a,}

{Improved air quality} {ISTEA, Environmental effects of

transportation decisions} {Carbon monoxide}

{ABC county} {Expand bus service}

Page 15: A planning methodology for intelligent urban transportation systems

Planning methodology for urban transportation systems 211

AND

THEN

Associated actions selected are

A, = {aj$ Use models M = {mili} with

Data D = {djli}

{HOV lanes, Advanced public transportation systems ( Expand bus service}

{Run models XYZ using data on transit statistics and origin destination from ABC county}

A complete set of rules is being developed for various models and data.

There are several challenges in integrating existing models. First, the current operational models often do not provide outputs that are useful for planning purposes. For example, the benefits of ATIS cannot be compared directly with those of HOV lanes using currently available models. A methodology for translating model inputs into a common comparable metric must be developed. Second, all existing transportation software has been developed independently with no regard to compatibility among models. Most transportation analysis models differ from each other in terms of their inputs, outputs, and internal structure. lntegrating them will require common commands and data formats. Finally, the outputs need to be translated into a form that is consistent with the PLANiTS environment and subsequent use in other models.

PLANiTS also contains an array of generic models such as regression and simulation to support higher levels of modeling intensity. The generic models will be used by people expe- rienced in quantitative analysis in instances where the knowledge base and the existing models do not adequately address a problem. A description of the generic models will be provided along with their appropriateness to specific applications. For example, regression is appropriate for analyzing continuous dependent variables such as incident durations, discrete choice analysis can be used to understand human decision making, and neural networks can be used for exploring functional relationships between inputs and outputs. Users can combine generic models to obtain richer insights. For example, discrete choice analysis can be used to estimate parameters of a disaggregate choice model that relates route selection to system performance, informaticn, socioeconomic and contextual factors. These parameters can then be used in a queuing model to assign traffic in a network. Users will be warned when greater complexity in analysis does not lead to better or more robust conclusions.

There will be rules to help the user find the appropriate model. An example of such rules would be:

Rule Simplified example

IF AND

AND

AND AND AND

THEN

AND

Goal is G = {gi} Policy factor of interest is

p = {Pi) Measures of Performance

y = {YJ Environment is E = {e,} Action selected is A = {ai} Associated action selected is

Aa = {aj,J Factors which influence {y,} are

F = {fi}

Use models M = {mjli} with Data

D = {dilJ

{Reduce congestion} {ISTEA, Congestion relief}

{Incident delay on freeway UVW and arterial XYZ)

{ABC links} {Incident management} {Advanced traveler information system (

Incident management} {factors influencing incident delay are:

percent of travelers equipped with information devices and network characteristics such as regular and incident reduced capacities of routes, free flow travel times, incident location and flow rate}

{deterministic queuing model, with data on incident durations, percent of travelers equipped with information, regular and incident capacities of routes, free flow travel times, incident location and flow rate}

TR(C) 2:4-B

Page 16: A planning methodology for intelligent urban transportation systems

212 A. KANAFANI er 01.

4. INTEGRATED PLANNING AND ANALYSIS PROCESS

At the heart of the methodology is a computer-based intelligent facilitator and decision support system to facilitate interactions among the many actors in the urban transportation planning process and the analytical and information resources with PLANiTS. It allows par- ticipants to work with the system and each other in an interactive online environment to perform analyses, deliberate, synthesize positions, and seek consensus. Alternative assumptions and goals as well as alternative actions can be examined. By means of its rich knowledge base and powerful analytic tools this computer-based intelligent facilitator is used to discover win-win propositions, to clarify trade-offs in meaningful, and when possible, quantitative terms, and to support trade-off analysis whenever optimal solutions are not possible.

4.1 The Planning Vector The central subject in the deliberative planning and analysis process supported in the

PLANiTS environment is the Planning vector PV, which is a symbolic representation of the central purpose of the planning activity. It contains three sub-vectors. The first is the Action vector A, which contains the proposed set of actions that are the subject of the planning process. The second is the Criteria vector Y, which contains the measures of performance representing the goals for which the actions are proposed and the criteria on the basis of which they are evaluated. The third is the Environment vector E, which contains the descriptors of the envi- ronment of the system at hand that are relevant to the subject actions. The essence of the deliberative process in PLANiTS is: (i) the construction of the Planning vector PV, (ii) use of the knowledge and the methods bases to analyze its contents, and (iii) use of the results to inform the decision-making process involved in programming projects in A.

The deliberative process may begin with any of the three elements of the PV. In trans- portation planning projects are often generated exogenously by a mandate, a funding opportu- nity, or a local deliberative process in which a community develops an action agenda and proposes it at a metropolitan or regional planning level. In such a case, the starting point is the action vector A or some elements of it. In other cases, a set of goals and objectives are used to drive the search for planning actions in a given environment. Therefore, the starting point is the criteria vector Y. An example follows:

An action a, is proposed-it could be a physical project, an operational improvement, or a control measure. The process begins with deliberation using the Strategy and Goals Base to determine the goals that the action is intended to address. This would result in selecting an initial set Y of performance criteria. At this point PLANiTS generated alternatives to the proposed action by exploring the Strategy and Action Base to determine what other actions, aJli address the same goals. A matrix such as that shown in Table 1 is interrogated to find actions that might complement or enhance the effectiveness of the proposed action or the alternatives. This search provides an opportunity to explore the potential of new IVHS technologies to enhance more conventional transportation projects. If the proposed action is an HOV lane segment in a given freeway corridor:

a, = {HOV Lane} the goal might be congestion reduction and therefore,

yI = {Delay) and the vector of associated actions A, might be:

ajl, = {Park-and-ride lots, Advanced rideshare matching program ( HOV lane} this results in a revised set of performance measures that now includes vehicle trip reduction:

yz(ajl, = {Delay, Number of vehicle trips} An alternative to the proposed action that would address the revised goals can be:

a 2 = {congestion pricing implemented through automatic vehicle identification system} and the vector of associated actions might be:

ajlz = {Advanced transit information systems, Teleworking, Increased downtown parking fees (to reduce downtown congestion) 1 Congestion pricing} this may result in further revisions of the performance measures:

yk/ajli = {Delay, Number of vehicle trips, Transit cost-effectiveness} Deliberation continues with these expanded action vectors; while the evaluation criteria for each

Page 17: A planning methodology for intelligent urban transportation systems

Planning methodology for urban transportation systems 213

set are reviewed in order to seek consensus on sets Y that users want to consider. The various components of this process are supported by the Strategy and Action Base, which calls on rules to inform the choice of alternate and complementary actions. At the same time, the case and

expert bases are used to check for consistency in the match between the elements of A and Y for the given system as defined by vector Y. The Strategy and Action Base, the Policy and Goals Base, and the Knowledge Base are all organized using a common object-oriented structure for storing information, case histories, and expert system rules.

The deliberative planning and action module function of the process supports brain- storming, goal setting, problem definition, and generation of alternatives using techniques categorized as team and task support systems (Vlahos, Khattak, Kanafani, and Manheim, 1994). The other function, consensus building and resolution seeking, supports the weighing of trade-offs, making of compromises, compensation of losers, and final decision making. There is no clear-cut division between these functions, the process does not proceed linearly, and there will be considerable back-and-forth before final decisions are made.

There is significant current research in computer support for such group process, and PLANiTS will adopt or develop appropriate support systems. PLANiTS will probably include the following functions (Vlahos, Manheim, Xie, 1994):

Information-sharing tools such as electronic brainstorming for generating ideas, threaded discussions, group dictionary, and collaborative writing Group evaluation functions including classification and categorization of issues, ranking, voting, and scoring Communications functions, which include electronic mail, chatting, and remote user access Meeting management tools including access control, team definition/assignment, deadline

management Personal work support, for example, links to various software packages Intelligent agents such as automatic mail processing and rule-based routing of information.

The development of team and task support mechanisms will allow multiple users to work together by exchanging information that includes components of the planning vector.

4.2 The PLANiTS planning environment We envision PLANiTS having many users: transportation professionals, decision makers,

citizens’ groups, and individual citizens, at the city, county congestion-management agency, and regional-planning agency levels. Users may begin with a diagnostic tour that would consist of graphic representation of congestion, delay, emissions, noise, and accidents, by transporta- tion corridor, geographic location, mode, time of day, day of week, and time of year. The tour is intended to inform users about the temporal patterns of congestion, the location and duration of bottlenecks, the occurrence of incidents, and other relevant aspects of the environment. Thereby, users would gain an understanding of the problems facing the region, the source of the problems, and potential solutions.

Working individually or in groups, PLANiTS users will interact with the PLANiTS Policy and Goals base, Strategy and Action base, and Data and Knowledge base and with each other to develop the elements of the planning vector. They may approach the analysis by proposing a particular action or a set of actions or by posing a problem or group of problems. Searching through the case base and the expert base, PLANiTS will use matching and pattern recognition to support analysis of proposed planning vectors with similar cases and knowledge from ex- perts. The expert system will advise the users on the adequacy of available knowledge, the need for primary data collection, and the appropriate selection of models for evaluating the proposed planning actions. By placing the planning process in a computer-aided environment, PLANiTS aims to make it more transparent so users can see the effects of differences in assumptions, criteria, and models of analysis.

Electronic brainstorming and electronic commenting will be used to generate and refine ideas and programs during this process. Areas of dominance will be identified and described, as well as situations where a trade-off is necessary. Users will deliberate regarding the relative importance of various goals, the validity of assumptions, the cost-effectiveness of various

Page 18: A planning methodology for intelligent urban transportation systems

214 A. KANAFANI et al.

improvements, and the timing of improvements. PLANiTS will contain models of multicriteria evaluation and decision making that can assist users by synthesizing the deliberations and highlighting the important aspects of the negotiation. Considerable feedback will probably occur at this stage. as the planners go back through PLANiTS to revise certain elements of the planning vector in order to explore the impacts of different actions or policies. The use of PLANiTS does not guarantee consensus, especially if there is significant conflict between the interests of various groups. However, by clarifying goals and exposing differences in interests, it could facilitate a more satisfactory compromise between groups than otherwise possible.

6. SUMMARY AND CURRENT RESEARCH AGENDA

We have proposed a new, computer-aided environment for urban transportation planning and a methodology that integrates analysis and deliberation. Key support features are:

0 Use of previously accumulated knowledge, appropriately synthesized, to address the plan- ning and policy issues at hand

0 Integration of models in a unified framework along with tools for developing new models 0 Use of tools that support group processes such as deliberation and brainstorming 0 Provision of intelligent advice when requested.

The methodology is designed to facilitate the entire planning and programming process at the local, regional, and state levels. Furthermore, interested individuals and/or groups could also use PLANiTS to identify problems and address them in a comprehensive manner through analysis and judgement .

We are proceeding with the development of a prototype and some important components. An important aspect of this development is to look for ways in which the planning system can be made more useful to the political processes of planning. Clearly PLANiTS must have political support if it is to be of value in the real world of decision making and transportation programming. With this in mind, the following research activities are underway.

The prototype has a complete shell of the overall structure of PLANiTS that can be demonstrated on a limited scale in a desktop-computer environment. The prototype includes a few planning vectors intended for illustrative purposes. It is integrating some current planning and operations models and developing the bases. The knowledge base is perhaps the most unique feature of the proposed methcdology. We are developing a framework for knowledge representation in transportation planning, and combining this with adaptations of techniques such as case-based reasoning and data filtering to create a knowledge base for planning. We are selecting the actions programmed in the PLANiTS prototype to begin to enrich the knowledge base with real information.

We are developing a platform for the integration of planning and operations models into a methods base for PLANiTS. Also, a large effort is underway to develop models of traveler and system behavior under IVHS scenarios. We are hoping that the PLANiTS environment becomes a unifying platform for such model integration activities. Finally, we are continuing to specify the requirements of a decision support environment for PLANiTS.

Acknowledgement-This research was conducted at the Institute of Transportation Studies of the University of California at Berkeley. It was funded by the California Department of Transportation, as a part of the PATH Program. We are grateful to the anonymous reviewers for their input.

REFERENCES

CBR (1988) Proceedings, Case-Based Reasoning Workshop, Clearwater. FL, May 10-13. Sponsored by Defense Advanced Research Projects Agency (DARPA), Information Science and Technology Office, FL.

CBR (1989) Proceedings. Case-based Reasoning Workshop. Pensacola Beach, FL. May 31-June 2. Sponsored by DARPA, Information Science and Technology Office, FL.

Guariso G. and Werthner H. (1989) Environmental Decision Support Systems. Ellis Horwood Limited, West Sussex, England.

Kanafani A, Khattak A., Crotty M.. and Dahlgren J. (1993) A planning methology for intelligent urban transportation

Page 19: A planning methodology for intelligent urban transportation systems

Planning methodology for urban transportation systems 215

systems. Partners in Advanced Transit and Highways (PATH) Research Report UCB-ITS-PRR-93-14. Institute of Transportation Studies, University of California at Berkeley, CA.

Kolodner J. (1993) Case-Based Reasoning. Morgan Kaufmann Publishers, San Mateo, CA. OECD (1990) OECD Workshop on Knowledge-based Expert Sysretns in Transportation. Espoo, Finland, 26-28 June

1990. Sponsored by the Organisation for Economic Cooperation and Development, Finland. Ritchie S. (1987) Expert systems in pavement management. Transportation Research. 21A, 145-152. Tung R. and Schneider J. (1987) Designing optimal transportation networks: An expert systems approach. Transpor-

t&on Research Record, 1145, 20-27. Varaiya P. and Shladover S. (1991) Sketch of an IVHS systems architecture. PATH Research Report UCB-ITS-PRR

91-3. Institute of Transportation Studies, University of California at Berkeley, CA. Vlahos N., Khattak A., Kanafani A., and Manheim M. (1994) The role of team work in a planning methodology for

intelligent transportation systems. Transporrarion Research, 2C. 217-229. Vlahos N, Manheim M., and Xie Y. (in preparation) Design of a support infrastructure for a transportation planning

and programming process. PATH Research Report. Institute of Transportation Studies, University of California at Berkeley, CA.