U.S. Department of Transportation
Federal Highway Administration
1200 New Jersey Avenue, SE
Washington, DC 20590
202-366-4000
Federal Highway Administration Research and Technology
Coordinating, Developing, and Delivering Highway Transportation Innovations
This report is an archived publication and may contain dated technical, contact, and link information |
|
Publication Number: FHWA-RD-02-045
Date: March 2003 |
||||||||||||||||
IHSDM Intersection Diagnostic Review Model1.0 INTRODUCTIONThis report documents the results of Federal Highway Administration (FHWA) Contract No. DTFH61-97-C-00053, "Development of an Expert System for the Interactive Highway Safety Design Model (IHSDM)." The objective of this contract was to develop software to perform a diagnostic review of intersections on rural two-lane highways, referred to as the Intersection Diagnostic Review Model (IDRM). The two main components of the project are:
This report focuses on documenting the knowledge base developed for IDRM. Prototype software implementing this knowledge base was completed and delivered as part of the project. The software is documented as part of the overall IHSDM software documentation. This report also documents the software in that it identifies the knowledge structure, problem definitions, models, decision algorithms, formulas, and parameter values implemented in the software. This introduction summarizes the objectives, scope, and approach of IDRM, and its role in IHSDM. 1.1 Role of the Intersection Diagnostic Review Model (IDRM) in the IHSDMThis project is part of the FHWA's Interactive Highway Safety Design Model (IHSDM). IHSDM is an integrated suite of tools to assist highway designers in evaluating and improving the consideration of safety in their designs. IHSDM is being developed in a modular manner and includes several analysis modules, including:
The initial priority in IHSDM development is being given to the design of projects on rural two-lane highways. Within that scope, IDRM focuses on the design of intersections on rural two-lane highways. The role and function of IDRM within IHSDM are illustrated in Figure 1 . An IDRM review of an intersection design typically would be performed after the application of the IHSDM Crash Prediction, Policy Review, and Design Consistency Modules, which generate output that may be used as input to IDRM. IDRM operates within the IHSDM user interface and obtains computer-aided design (CAD) data through the IHSDM CAD interface. The intended user for IDRM is the working highway designer or traffic engineer. IDRM will be used at most stages of the design process, including engineering studies, functional/conceptual design, preliminary design, and detailed design. For example, an engineer might use IDRM to review a conceptual design option for a study to ensure that it does not contain combinations of geometric features that would require changing the concept at a later stage of design. 1.2 IDRM ObjectivesThe objective of IDRM is to supplement policy and standards by providing a comprehensive diagnostic review of an intersection design, analogous to a review that might be performed in a design organization by a senior highway safety engineer. A primary focus is to identify combinations of geometric design elements that suggest potential concerns, even though the elements individually would be considered within acceptable practice. A number of potential issues were identified as candidates to be addressed by IDRM; selection of the issues incorporated in IDRM is discussed in Section 2.1. IDRM was conceived as an expert system – i.e., a software system that emulates the knowledge of a human expert. An expert system is clearly suggested by the objective of providing a function similar to an expert design review. Expert systems differ from other software in that their intent is to emulate expert reasoning. They typically provide greater depth and flexibility in representing knowledge. The starting points for an expert system are:
These were the overall tasks in developing the IDRM knowledge base. A second expert system, referred to as the Project Development Expert System, was included in the scope of this project. Whereas the intent of the IHSDM Expert System is to review a proposed or potential highway design, the Project Development Expert System is intended as a tool to screen existing intersections for improvements. During the course of the project, the research team concluded that the same knowledge and decision process is applicable to both the IHSDM Expert System and the Project Development Expert System. Therefore, it was decided to develop a single, unified knowledge base and software product for both systems. A key distinction between the IHSDM Expert System and the Project Development System is that crash history data for an existing intersection may be available, whereas such data will not be available for a proposed new design. (Crash data may, however, be available for an actual intersection design and traffic conditions similar to a new design.) Such data may be of value in assessing the need for design improvements. (However, crash frequency may often be too low to be statistically significant, especially for rural two-lane highways.) Accordingly, the unified expert system includes the capability to apply crash data where available, as described in Section 5.0. 1.3 The Knowledge Base Development ProcessMembers of the IDRM research team are recognized experts in highway design and highway safety research. The team also included expert system development specialists with experience in the process of capturing human expert knowledge into software – "knowledge engineering." In consultation with FHWA, it was decided that the IDRM knowledge base should be derived primarily from the experts who were included in the research team for that purpose, together with relevant published literature and consultation with specialists where appropriate. A relatively small number of experts who are available for extended periods of time greatly facilitate the process of knowledge engineering. To obtain a broader range of input and to help validate the knowledge base, an Advisory Panel was created. The panel consisted of representatives of six state highway departments, as well as FHWA representatives. The panel met four times over the course of the project to review the knowledge base and to provide general guidance. Advisory Panel participants are listed in Table 1 . The research team wishes to acknowledge the Advisory Panel’s substantial contribution to the project. The knowledge base development process consisted of the following steps:
Table 1. IDRM Advisory Panel Participants
1.4 IDRM as an Expert SystemExpert systems are intended to "behave" in a manner similar to a consultation with a human expert. To accomplish these goals, expert systems usually use search algorithms to locate and apply knowledge, rather than the sequential execution of program statements of "procedural" software. As discussed in Section 1.3, Expert systems are software tools that are used to capture and make accessible a body of human knowledge that is capable of being codified in a logical structure, such as rules. This body of knowledge is referred to as the "knowledge base." Expert systems are intended to "behave" in a manner similar to a consultation with a human expert. To accomplish these goals, expert systems usually use search algorithms to locate and apply knowledge, rather than the sequential execution of program statements of "procedural" software. As discussed in Section 1.5, the IDRM knowledge base is composed mainly of engineering models that quantitatively address the concerns IDRM is designed to evaluate. Rule-like structures are used to select the appropriate model for a given situation and to generate model inputs. Elements of IDRM design and operation that reflect the expert system approach include:
Figure 1. IHSDM Functional Overview 1.5 IDRM Knowledge BaseIDRM, like all expert systems, is developed from a base of expert knowledge concerning the issue at hand – in this case, the design of at-grade intersections on rural two-lane highways. The knowledge base consists of a set of decision rules for determining whether or not each of a set of defined potential design concerns is present. IDRM identifies a level for each concern found to be present: Level 1 – Concern could indicate a potential safety issue. Level 2 – Concern could indicate the potential for significant design improvement. It is emphasized that the basis for identification of a concern by the knowledge base is the presence of an issue that should be considered by the designer. A concern identified by the knowledge base does not necessarily indicate a problem or a need to change the design. The knowledge base is deliberately designed to be conservative in suggesting concerns to the user, even in cases where no problem may be present. Designers are generally knowledgeable about the basic geometric and traffic control requirements for intersection design. Most design agencies have well-developed procedures and standards for executing "typical" designs. The problems that emerge, and those for which many designers have little current guidance, involve combinations of geometric problems or unusual conditions in conjunction with the intersection. Many design concerns may be recognized as being important, yet they tend to be treated qualitatively, and hence may not always be addressed by designers. Thus, IDRM focuses on identification of both specific individual problems and combinations of problems related to geometric design. Whereas many expert systems consist of networks of "if-then" rules, the primary element of the IDRM knowledge base is a set of engineering models. Each potential concern is evaluated by using an engineering model to calculate one or more evaluation parameters. The appropriate model is selected based on the intersection geometry and the concern being evaluated. The user is queried for any information needed by the model that is not available in the design information input. The parameter is then compared to threshold values to determine the presence and level of the concern. These models are presented in detail in Section 3.0 . IDRM also identifies potential design improvements – "treatments" – that could be considered to address each concern. Potential treatments to address concerns are identified by decision rules in the form of a matrix that relates combinations of concerns and geometric elements to treatments. Candidate treatments are divided into "design improvements," which involve some change to the highway design, and "mitigation measures," which can be implemented without a design change. The treatment decision rules are documented in Section 6.0 . These decision rules, models, and threshold values comprise the IDRM knowledge base. 1.6 Use of IDRM ResultsIDRM is intended for use by the working highway designer or traffic engineer. IDRM can be used at most stages of the design process, including engineering studies, functional/conceptual design, preliminary design, and detailed design. For example, an engineer might use IDRM to review a conceptual design option for a study to ensure that it does not contain combinations of geometric features that would require changing the concept at a later stage of design. A concern identified by IDRM represents an issue that should be considered by the designer. It does not necessarily indicate a problem or a need to change the design. The knowledge base is deliberately designed to be conservative in suggesting concerns to the user, even in cases where no problem may be present. Level 1 concerns represent potential safety issues and should be carefully considered by the designer. Level 2 concerns represent potential opportunities for significant design improvement. IDRM usually identifies multiple potential mitigation measures for identified concerns, with an impact on the intersection design ranging from minor additions or revisions to major geometric redesign. This is intended to give the designer multiple options to address the concern. It is expected that the designer will consider the stage of design, cost, and geometric and other
constraints in selecting an appropriate mitigation measure.
|