Operational Level of Service Analysis Using Probe Vehicle Data

Currently, the Highway Capacity Manual (HCM) method for operational level of service analysis is dependent on traffic demand counts that are captured in the field as departure volumes and queue counts. On freeways, capturing queue counts in the field poses worker and driver safety hazards. Recent studies explored the use of unmanned aerial vehicles to capture freeway queueing; however, range and battery life presented a challenge. In addition, conventional and novel methods are not cost effective and require considerable amount of time for planning the work, field data collection, and performing the analysis. Furthermore, the current state-of-the-practice limits analysts to few or even one traffic conditions to be analyzed. This yields to one of the most pressing issues in the field of highway operations being the lack of a cost-effective retrospective method to perform level of service analysis. The issue is problematic because it affects how agencies align their investments for congestion relief and safety improvement projects with various traffic conditions. The issue persists because little is known about the statistical correlation between level of service and readily available year-round data such as travel time from the National Performance Management Research Data Set (NPMRDS). In addition, current HCM methods rely on field collected data as explanatory variables. The purpose of this study is to (a) explore the dependency of level of service on congestion as expressed by travel time index and speed standard normal deviate, and (b) develop a representative relationship between level of service a function of the aforementioned congestion indices. This study was conducted using a generalized linear modeling approach. Models were developed using backward elimination while maintaining goodness of fit by means of the Bayesian Information Criterion. Probe-vehicle data and traffic counts from Texas were used to develop and validate the models. Key findings include pioneer statistically representative models that quantify level of service as explained by congestion indices. The study is significant for linking data-driven operational analyses and decision-making about aligning congestion relief and safety improvement investments.