DEVELOPMENT OF AN EXPERT SYSTEM TO ASSIST THE INTERACTIVE GRAPHIC TRANSIT SYSTEM DESIGN PROCESS.

摘要

Urban public transit has a vital role in the functioning of most urban areas. Transit network design and evaluation are important components of urban transit planning. Because of their complexity, they can be best handled using interactive graphic methods, coupled with ways to evaluate alternative transit network designs with respect to multiple, conflicting criteria. This dissertation research is focused on an investigation of the applicability of a knowledge-based expert system approach to increasing the productivity of the transit network design process.; For this research, an interactive knowledge-based expert system (TNOP-ADVISOR) has been developed to assist the development of high performance transit network designs. TNOP-ADVISOR provides advice about how to modify designs so as to obtain improved performance. A network simulation software package (TNOP) provides the capability for modifying and predicting the performance of these designs. The knowledge-based inference engine is capable of generating advice about what operational and system planning design changes are likely to lead to higher performance levels. Operational variables include headways, vehicle types, layover and departure times. System planning variables include route layout changes or adding/deleting entire routes. The knowledge-base can be applied to any transit network design problem at any stage of the design process and also can give advice irrespective of the demand pattern or base network being examined. A computer-based interactive multicriteria evaluation method that uses concordance analysis is developed and used to evaluate and rank the alternative transit system designs. It includes a non-linear method of normalization that has not been available before.; Three tests are conducted using the interactive knowledge-based expert system and multicriteria evaluation capability as support tools for the conduct of a TNOP-based design process. The results obtained show that the advice provided helped produce high performance designs in all cases. Human judgement is still required in some cases to translate general advice into specific transit system changes but advice on operational changes is specific enough to use directly. This system represents a first attempt to develop and test an intelligent CAD system for use in this field.