Optimization Modeling for Urban Street Design

Abstract

For decades transportation legislation actions have demonstrated the desire to plan, design and operate multi-modal surface transportation systems (National Complete Streets Coalition [NCSC], 2009). The push for multi-modal operations stems from several key concerns including environmental impacts, natural resource scarcity, rising fuel costs and dependency on foreign oil, and the declining health of Americans due to their reliance on personal automobile travel. The introduction of legislation for multi-modal surface transportation designs reflects the desire of the public and decision makers to provide greener designs that reduce our dependency on foreign oil and effects on the environment while improving air quality and the health of travelers. However, it has been determined that the methods needed by engineers and planners to design such facilities are currently lacking in their ability to reflect traveler perceptions of service by mode which is needed to successfully design such multi-modal transportation systems. In addition, design guidance does not include methods by which engineers and planners can weigh the range of potential alternative designs to optimize the design of streets to comfortably accommodate all modal travelers.

The purpose of this dissertation was to develop a Multi-objective Optimization Model to support the design of Complete Streets and to identify optimal urban street designs that achieve a pre-defined level of service rating for travelers on an urban arterial including auto, pedestrian and bicycle modal users, while meeting geometric design standards. To achieve this goal, Cumulative Logit Level of Service (LOS) Models were developed for the pedestrian and bicycle modes that incorporate traveler’s perceptions of Level of Service and provide a distribution of perceived LOS to assist decision makers. Next, a Multi-objective Optimization Model was developed that can provide an optimal right of way design to accommodate the auto, pedestrian and bicycle modes at a pre-defined LOS that also adhere to geometric design standards.

Building on a national research study database, the probabilities of road user perceptions of Level of Service (LOS) for the pedestrian and bicycle modes were developed using the Cumulative Logit Modeling technique. An existing auto cumulative logit LOS model was utilized and the transit mode was not included due to lack of similar data. These models used variables found to be statistically significantly correlated to traveler’s perception of LOS including: Space Mean Speed and Median Presence for the auto mode; Number of Traffic Lanes, and Sidewalk Width for the pedestrian mode and Number of Traffic Lanes, Bike/Shoulder Width and Posted Speed Limit for bicycle mode. These newly developed Cumulative Logit LOS Models for the pedestrian and bicycle modes provide a distribution of LOS ratings based on traveler perceptions of LOS and require minimal data collection on the part of the engineer or planner without a significant reduction in model accuracy which should spur the use of the methodology.

Next, these Cumulative Logit LOS Models were used in the development of a four-step Multi-objective Optimization Model that provides designers with a set of urban street characteristics that optimize modal traveler perceptions of service. The objective function was to balance the perceived LOS for each of the three modes subject to street characteristic constraints. Several scenarios of Right of Way (ROW) width for the Multi-objective Optimization Model were created to demonstrate the usefulness of the modeling approach. It was observed that fewer number of through lanes and the presence of a raised median, sidewalk and bike lane result in higher user rating of LOS for all three modes. These findings reflect the findings of previous studies conducted in-vehicle, through surveys, and through focus groups (Pecheux et al., 2004; Petritsch et al., 2005).

The findings of this study support the use of Cumulative Logit Modeling techniques to model ordered categorical traveler perception LOS data with a reduced set of independent variables for the pedestrian and bicycle modes on urban streets as compared to previous data intensive models (Dowling (NCHRP report 616). In addition, this study provides a new method for designing Complete Streets that seeks to optimize the perceived performance of urban streets by the auto, pedestrian, and bicycle modes while adhering to existing design standards.